*/}
{/* {npzData.map((npz) => (
diff --git a/react-ui/src/tabs/BarkGenerationParams.tsx b/react-ui/src/tabs/BarkGenerationParams.tsx
index 4c676c2f..f85db914 100644
--- a/react-ui/src/tabs/BarkGenerationParams.tsx
+++ b/react-ui/src/tabs/BarkGenerationParams.tsx
@@ -1,31 +1,18 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
+import { parseFormChange } from "../data/parseFormChange";
+import { generateWithBark } from "../functions/generateWithBark";
+import { useHistory } from "../hooks/useHistory";
import { BarkResult } from "./BarkResult";
+import { getBarkFuncs } from "../data/getBarkFuncs";
+import { Seeded } from "../types/Seeded";
-const inputs = {
- burn_in_prompt: "Howdy!",
- prompt: "Howdy!",
- history_setting: "Empty history",
- languageRadio: "English",
- speakerIdRadio: "0",
- useV2: true,
- text_temp: 0.7,
- waveform_temp: 0.7,
- long_prompt_radio: "Short prompt (<15s)",
- long_prompt_history_radio: "or Use old generation as history:",
- old_generation_dropdown:
- "voices\\2023-06-18_21-51-07__bark__continued_generation.npz",
- seed: "123",
- history_prompt_semantic_dropdown:
- "voices\\2023-06-18_21-51-07__bark__continued_generation.npz",
- max_gen_duration_s: 15,
-};
-
-export type BarkGenerationParams = {
+export type BarkGenerationParams = Seeded & {
burn_in_prompt: string; // string in 'Burn In Prompt (Optional)' Textbox component
- prompt: string; // string in 'Prompt' Textbox component
+ text: string; // string in 'Prompt' Textbox component
history_setting: string; // string in 'History Prompt (voice) setting:' Radio component
languageRadio: string; // string in 'parameter_17' Radio component
speakerIdRadio: string; // string in 'Speaker ID' Radio component
@@ -35,16 +22,30 @@ export type BarkGenerationParams = {
long_prompt_radio: string; // string in 'Prompt type' Radio component
long_prompt_history_radio: string; // string in 'For each subsequent generation:' Radio component
old_generation_dropdown: string; // string
- seed: string; // string in 'parameter_40' Textbox component
history_prompt_semantic_dropdown: string; // string
max_gen_duration_s: number; // number in 'Max generation duration (s)' Number component
};
export const initialState: BarkGenerationParams = {
- ...inputs,
+ seed: 0,
+ use_random_seed: true,
+
+ burn_in_prompt: "",
+ text: "",
+ history_setting: "Empty history",
+ languageRadio: "English",
+ speakerIdRadio: "0",
+ useV2: true,
+ text_temp: 0.7,
+ waveform_temp: 0.7,
+ long_prompt_radio: "Short prompt (<15s)",
+ long_prompt_history_radio: "or Use old generation as history:",
+ old_generation_dropdown: "",
+ history_prompt_semantic_dropdown: "",
+ max_gen_duration_s: 15,
};
-export const barkGenerationId = "bark_generation-tab.v2";
+export const barkGenerationId = "bark_generation-tab.v3";
export const sendToBarkAsVoice = (old_generation_dropdown?: string) => {
if (!old_generation_dropdown) return;
@@ -65,5 +66,29 @@ export const sendToBarkAsVoice = (old_generation_dropdown?: string) => {
export const useBarkGenerationParams = () =>
useLocalStorage(barkGenerationId, initialState);
-export const useBarkResult = () =>
- useLocalStorage(barkGenerationId + ".output", null);
+export const getBarkGenerationParams = (): BarkGenerationParams =>
+ readLocalStorage(barkGenerationId) ?? initialState;
+
+export const useBarkPage = () => {
+ const [barkGenerationParams, setBarkGenerationParams] =
+ useBarkGenerationParams();
+
+ const [historyData, setHistoryData] = useHistory("bark");
+
+ async function consumer(params: BarkGenerationParams) {
+ if (params.use_random_seed)
+ setBarkGenerationParams((x) => ({ ...x, seed: params.seed }));
+ const result = await generateWithBark(params);
+ setHistoryData((historyData) => [result, ...historyData]);
+ }
+
+ return {
+ barkGenerationParams,
+ setBarkGenerationParams,
+ historyData,
+ setHistoryData,
+ consumer,
+ handleChange: parseFormChange(setBarkGenerationParams),
+ funcs: getBarkFuncs(setBarkGenerationParams, barkGenerationParams),
+ };
+};
diff --git a/react-ui/src/tabs/BarkResult.tsx b/react-ui/src/tabs/BarkResult.tsx
index 09dab850..c6735988 100644
--- a/react-ui/src/tabs/BarkResult.tsx
+++ b/react-ui/src/tabs/BarkResult.tsx
@@ -3,7 +3,7 @@ import { GradioFile } from "../types/GradioFile";
export type BarkResult = {
audio: GradioFile;
npz: string;
- json_text: {
+ metadata: {
_version: string;
_hash_version: string;
_type: string;
diff --git a/react-ui/src/tabs/DemucsParams.tsx b/react-ui/src/tabs/DemucsParams.tsx
index 1718c8e9..6d6e6309 100644
--- a/react-ui/src/tabs/DemucsParams.tsx
+++ b/react-ui/src/tabs/DemucsParams.tsx
@@ -1,5 +1,11 @@
-import { updateLocalStorageWithFunction } from "../hooks/useLocalStorage";
+import useLocalStorage, {
+ readLocalStorage,
+ updateLocalStorageWithFunction,
+} from "../hooks/useLocalStorage";
import router from "next/router";
+import { GradioFile } from "../types/GradioFile";
+import { splitWithDemucs } from "../functions/splitWithDemucs";
+import { useHistory } from "../hooks/useHistory";
export type DemucsParams = {
file?: string;
@@ -9,6 +15,26 @@ export const initialState: DemucsParams = {
file: "https://www.mfiles.co.uk/mp3-downloads/gs-cd-track2.mp3",
};
+export type DemucsGradioFile = GradioFile & {
+ type_name?: string;
+};
+
+type DemucsResult = {
+ audio: DemucsGradioFile;
+ typeName: string;
+};
+
+const typeNames = ["drums", "bass", "other", "vocals"];
+
+const addTypeNamesToAudioOutputs = (
+ audioOutputs: DemucsGradioFile[],
+ typeNames: string[]
+) =>
+ audioOutputs.map((audioOutput, index) => ({
+ ...audioOutput,
+ type_name: typeNames[index],
+ }));
+
export const demucsId = "demucs-tab";
export const sendToDemucs = (file?: string) => {
@@ -20,3 +46,44 @@ export const sendToDemucs = (file?: string) => {
);
router.push("/demucs");
};
+
+export const useDemucsParams = () =>
+ useLocalStorage(demucsId, initialState);
+
+export const useDemucsResults = () =>
+ useLocalStorage(demucsId + ".output", null);
+
+export const getDemucsParams = (): DemucsParams =>
+ readLocalStorage(demucsId) ?? initialState;
+
+export const useDemucsPage = () => {
+ const [demucsParams, setDemucsParams] = useDemucsParams();
+
+ const [historyData, setHistoryData] = useHistory("demucs");
+
+ const consumer = async (params: DemucsParams) => {
+ const result = await splitWithDemucs(params);
+ const sampleWithTypeNames = addTypeNamesToAudioOutputs(result, typeNames);
+ const data = typeNames.map((typeName) => ({
+ audio: sampleWithTypeNames?.find((item) => item.type_name === typeName)!,
+ typeName,
+ }));
+ setHistoryData((historyData) => [...data, ...historyData]);
+ };
+
+ const funcs = {
+ useAsInput: (audio?: string) => {
+ if (!audio) return;
+ setDemucsParams({ ...demucsParams, file: audio });
+ },
+ };
+
+ return {
+ demucsParams,
+ setDemucsParams,
+ historyData,
+ setHistoryData,
+ consumer,
+ funcs,
+ };
+};
diff --git a/react-ui/src/tabs/MMSParams.tsx b/react-ui/src/tabs/MMSParams.tsx
index acc154e7..64cd2e4f 100644
--- a/react-ui/src/tabs/MMSParams.tsx
+++ b/react-ui/src/tabs/MMSParams.tsx
@@ -1,12 +1,14 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
import { GradioFile } from "../types/GradioFile";
+import { Seeded } from "../types/Seeded";
const MMS_ID = "MMSParams";
-export type MMSParams = {
+export type MMSParams = Seeded & {
text: string;
language: string;
speaking_rate: number;
@@ -15,6 +17,9 @@ export type MMSParams = {
};
export const initialMMSParams: MMSParams = {
+ seed: 0,
+ use_random_seed: true,
+
text: "",
language: "eng",
speaking_rate: 1.0,
@@ -51,3 +56,6 @@ export const useMMSParams = () =>
export const useMMSResult = () =>
useLocalStorage(MMS_ID + ".output", null);
+
+export const getMMSParams = (): MMSParams =>
+ readLocalStorage(MMS_ID) ?? initialMMSParams;
diff --git a/react-ui/src/tabs/MagnetParams.tsx b/react-ui/src/tabs/MagnetParams.tsx
index fa1ba10f..d54bfa95 100644
--- a/react-ui/src/tabs/MagnetParams.tsx
+++ b/react-ui/src/tabs/MagnetParams.tsx
@@ -1,15 +1,21 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
import { GradioFile } from "../types/GradioFile";
+import { parseFormChange } from "../data/parseFormChange";
+import { useHistory } from "../hooks/useHistory";
+import { generateWithMagnet } from "../functions/generateWithMagnet";
+import { favorite } from "../functions/favorite";
+import { Seeded } from "../types/Seeded";
+import { useSeedHelper } from "../functions/results/useSeedHelper";
export const magnetId = "magnetParams";
-export type MagnetParams = {
+export type MagnetParams = Seeded & {
model: string;
text: string;
- seed: number;
use_sampling: boolean;
top_k: number;
top_p: number;
@@ -38,12 +44,13 @@ export const initialMagnetParams: MagnetParams = {
decoding_steps_3: 40,
decoding_steps_4: 40,
span_arrangement: "nonoverlap",
+ use_random_seed: true,
};
export type MagnetResult = {
audio: GradioFile;
history_bundle_name_data: string;
- json: {
+ metadata: {
_version: string;
_hash_version: string;
_type: string;
@@ -81,3 +88,49 @@ export const useMagnetParams = () =>
export const useMagnetResult = () =>
useLocalStorage(magnetId + ".output", null);
+
+export const getMagnetParams = (): MagnetParams =>
+ readLocalStorage(magnetId) ?? initialMagnetParams;
+
+export const useMagnetPage = () => {
+ const [magnetParams, setMagnetParams] = useMagnetParams();
+
+ const [historyData, setHistoryData] = useHistory("magnet");
+
+ const consumer = async (params: MagnetParams) => {
+ const data = await generateWithMagnet(params);
+ if (params.use_random_seed)
+ setMagnetParams((x) => ({ ...x, seed: params.seed }));
+ setHistoryData((x) => [data, ...x]);
+ return data;
+ };
+
+ const funcs = {
+ favorite,
+ useSeed: useSeedHelper(setMagnetParams),
+ useParameters: (_url: string, data?: MagnetResult) => {
+ const params = data?.metadata;
+ if (!params) return;
+ setMagnetParams({
+ ...magnetParams,
+ ...params,
+ seed: Number(params.seed),
+ model: params.model || initialMagnetParams.model,
+ decoding_steps_1: params.decoding_steps[0],
+ decoding_steps_2: params.decoding_steps[1],
+ decoding_steps_3: params.decoding_steps[2],
+ decoding_steps_4: params.decoding_steps[3],
+ });
+ },
+ };
+
+ return {
+ magnetParams,
+ setMagnetParams,
+ historyData,
+ setHistoryData,
+ consumer,
+ handleChange: parseFormChange(setMagnetParams),
+ funcs,
+ };
+};
diff --git a/react-ui/src/tabs/MahaParams.tsx b/react-ui/src/tabs/MahaParams.tsx
index 004daa53..c3cdc3e3 100644
--- a/react-ui/src/tabs/MahaParams.tsx
+++ b/react-ui/src/tabs/MahaParams.tsx
@@ -1,26 +1,34 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
import { GradioFile } from "../types/GradioFile";
+import { parseFormChange } from "../data/parseFormChange";
+import { favorite } from "../functions/favorite";
+import { generateWithMaha } from "../functions/generateWithMaha";
+import { useHistory } from "../hooks/useHistory";
+import { useSeedHelper } from "../functions/results/useSeedHelper";
+import { Seeded } from "../types/Seeded";
const mahaId = "mahaParams";
-export type MahaParams = {
- maha_tts_input: string;
+export type MahaParams = Seeded & {
+ text: string;
model_language: string;
maha_tts_language: string;
speaker_name: string;
- seed: number;
device: string;
};
export const initialMahaParams: MahaParams = {
- maha_tts_input: "",
+ seed: 0,
+ use_random_seed: true,
+
+ text: "",
model_language: "Smolie-in",
maha_tts_language: "english",
speaker_name: "",
- seed: -1,
device: "auto",
};
@@ -53,3 +61,45 @@ export const useMahaParams = () =>
export const useMahaResult = () =>
useLocalStorage(mahaId + ".output", null);
+
+export const getMahaParams = (): MahaParams =>
+ readLocalStorage(mahaId) ?? initialMahaParams;
+
+export const useMahaPage = () => {
+ const [mahaParams, setMahaParams] = useMahaParams();
+
+ const [historyData, setHistoryData] = useHistory("maha");
+
+ const consumer = async (params: MahaParams) => {
+ const data = await generateWithMaha(params);
+ if (params.use_random_seed)
+ setMahaParams((x) => ({ ...x, seed: params.seed }));
+ setHistoryData((x) => [data, ...x]);
+ return data;
+ };
+
+ const funcs = {
+ favorite,
+ useSeed: useSeedHelper(setMahaParams),
+ useParameters: (_url: string, data?: MahaResult) => {
+ const params = data?.metadata;
+ if (!params) return;
+ setMahaParams({
+ ...mahaParams,
+ ...params,
+ seed: Number(params.seed),
+ });
+ },
+ };
+
+ return {
+ mahaParams,
+ setMahaParams,
+ resetParams: () => setMahaParams({ ...mahaParams, ...initialMahaParams }),
+ historyData,
+ setHistoryData,
+ consumer,
+ handleChange: parseFormChange(setMahaParams),
+ funcs,
+ };
+};
diff --git a/react-ui/src/tabs/MusicgenParams.tsx b/react-ui/src/tabs/MusicgenParams.tsx
index c4cf8d2a..4cb5dbf4 100644
--- a/react-ui/src/tabs/MusicgenParams.tsx
+++ b/react-ui/src/tabs/MusicgenParams.tsx
@@ -1,12 +1,19 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
import { GradioFile } from "../types/GradioFile";
+import { Seeded } from "../types/Seeded";
+import { useHistory } from "../hooks/useHistory";
+import { generateWithMusicgen } from "../functions/generateWithMusicgen";
+import { parseFormChange } from "../data/parseFormChange";
+import { useSeedHelper } from "../functions/results/useSeedHelper";
+import { favorite } from "../functions/favorite";
-export const musicgenId = "musicgenParams";
+export const musicgenId = "musicgenParams.v2";
-export type MusicgenParams = {
+export type MusicgenParams = Seeded & {
text: string;
melody?: string;
model: string;
@@ -15,29 +22,28 @@ export type MusicgenParams = {
topp: number;
temperature: number;
cfg_coef: number;
- seed: number;
use_multi_band_diffusion: boolean;
};
export const initialMusicgenParams: MusicgenParams = {
- // text: "",
+ seed: 0,
+ use_random_seed: true,
+
text: "lofi hip hop beats to relax/study to",
melody: undefined,
- // melody: "https://www.mfiles.co.uk/mp3-downloads/gs-cd-track2.mp3",
model: "facebook/musicgen-small",
duration: 1,
topk: 250,
topp: 0,
temperature: 1.0,
cfg_coef: 3.0,
- seed: -1,
use_multi_band_diffusion: false,
};
export type MusicgenResult = {
audio: GradioFile;
history_bundle_name_data: string;
- json: {
+ metadata: {
_version: string;
_hash_version: string;
_type: string;
@@ -74,3 +80,52 @@ export const useMusicgenParams = () =>
export const useMusicgenResult = () =>
useLocalStorage(musicgenId + ".output", null);
+
+export const getMusicgenParams = (): MusicgenParams =>
+ readLocalStorage(musicgenId) ?? initialMusicgenParams;
+
+export const useMusicgenPage = () => {
+ const [musicgenParams, setMusicgenParams] = useMusicgenParams();
+
+ const [historyData, setHistoryData] = useHistory("musicgen");
+
+ const consumer = async (params: MusicgenParams) => {
+ const data = await generateWithMusicgen(params);
+ if (params.use_random_seed)
+ setMusicgenParams((x) => ({ ...x, seed: params.seed }));
+ setHistoryData((x) => [data, ...x]);
+ return data;
+ };
+
+ const funcs = {
+ useAsMelody: (melody?: string, metadata?: MusicgenResult) => {
+ if (!melody) return;
+ setMusicgenParams({
+ ...musicgenParams,
+ melody,
+ });
+ },
+ favorite,
+ useSeed: useSeedHelper(setMusicgenParams),
+ useParameters: (_url: string, data?: MusicgenResult) => {
+ const params = data?.metadata;
+ if (!params) return;
+ setMusicgenParams({
+ ...musicgenParams,
+ ...params,
+ seed: Number(params.seed),
+ model: params.model || "facebook/musicgen-small",
+ });
+ },
+ };
+
+ return {
+ musicgenParams,
+ setMusicgenParams,
+ historyData,
+ setHistoryData,
+ consumer,
+ handleChange: parseFormChange(setMusicgenParams),
+ funcs,
+ };
+};
diff --git a/react-ui/src/tabs/RVCParams.tsx b/react-ui/src/tabs/RVCParams.tsx
index d587576a..c4c7afdc 100644
--- a/react-ui/src/tabs/RVCParams.tsx
+++ b/react-ui/src/tabs/RVCParams.tsx
@@ -1,7 +1,12 @@
+import { useHistory } from "../hooks/useHistory";
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
+import { RVCResult } from "./RVCResult";
+import { applyRVC } from "../functions/applyRVC";
+import { parseFormChange } from "../data/parseFormChange";
export type RVCParams = {
pitch_up_key: string; // string in 'Pitch Up key' Textbox component
@@ -47,3 +52,73 @@ export const sendToRVCGeneration = (audio?: string) => {
export const useRVCGenerationParams = () =>
useLocalStorage(RVCId, initialState);
+
+export const getRVCGenerationParams = (): RVCParams =>
+ readLocalStorage(RVCId) ?? initialState;
+
+export const useRVCPage = () => {
+ const [rvcGenerationParams, setRvcGenerationParams] =
+ useRVCGenerationParams();
+
+ const [historyData, setHistoryData] = useHistory("rvc");
+
+ const consumer = async (params: RVCParams) => {
+ const data = await applyRVC(params);
+ setHistoryData((x) => [data, ...x]);
+ return data;
+ };
+
+ const funcs = {
+ useParameters: (_url: string, data?: RVCResult) => {
+ const {
+ f0up_key: pitch_up_key,
+ index_path: index,
+ f0method: pitch_collection_method,
+ model_path: model,
+ index_rate: search_feature_ratio,
+ device,
+ is_half: use_half_precision_model,
+ filter_radius: filter_radius_pitch,
+ resample_sr: resample_sample_rate,
+ rms_mix_rate: voice_envelope_normalizaiton,
+ protect: protect_breath_sounds,
+ } = data?.metadata ?? {};
+
+ setRvcGenerationParams({
+ ...rvcGenerationParams,
+ pitch_up_key: pitch_up_key ?? rvcGenerationParams.pitch_up_key,
+ // original_audio,
+ index: index ? index.replace(".index", "") : rvcGenerationParams.index,
+ pitch_collection_method:
+ pitch_collection_method ??
+ rvcGenerationParams.pitch_collection_method,
+ model: model ? model.replace(".pth", "") : rvcGenerationParams.model,
+ search_feature_ratio:
+ search_feature_ratio ?? rvcGenerationParams.search_feature_ratio,
+ device: device ?? rvcGenerationParams.device,
+ use_half_precision_model:
+ use_half_precision_model ??
+ rvcGenerationParams.use_half_precision_model,
+ filter_radius_pitch:
+ filter_radius_pitch ?? rvcGenerationParams.filter_radius_pitch,
+ resample_sample_rate:
+ resample_sample_rate ?? rvcGenerationParams.resample_sample_rate,
+ voice_envelope_normalizaiton:
+ voice_envelope_normalizaiton ??
+ rvcGenerationParams.voice_envelope_normalizaiton,
+ protect_breath_sounds:
+ protect_breath_sounds ?? rvcGenerationParams.protect_breath_sounds,
+ });
+ },
+ };
+
+ return {
+ rvcGenerationParams,
+ setRvcGenerationParams,
+ historyData,
+ setHistoryData,
+ consumer,
+ handleChange: parseFormChange(setRvcGenerationParams),
+ funcs,
+ };
+};
diff --git a/react-ui/src/tabs/TortoiseGenerationParams.tsx b/react-ui/src/tabs/TortoiseGenerationParams.tsx
index 9d792730..a637a1d7 100644
--- a/react-ui/src/tabs/TortoiseGenerationParams.tsx
+++ b/react-ui/src/tabs/TortoiseGenerationParams.tsx
@@ -1,14 +1,20 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
import { TortoiseResult } from "./TortoiseResult";
+import { Seeded } from "../types/Seeded";
+import { parseFormChange } from "../data/parseFormChange";
+import { favorite } from "../functions/favorite";
+import { generateWithTortoise } from "../functions/generateWithTortoise";
+import { useSeedHelper } from "../functions/results/useSeedHelper";
+import { useHistory } from "../hooks/useHistory";
-export type TortoiseGenerationParams = {
- prompt: string; // string in 'Prompt' Textbox component
- speaker: string; // string (Option from: ['random', 'angie', 'applejack', 'cond_latent_example', 'daniel', 'deniro', 'emma', 'freeman', 'geralt', 'halle', 'jlaw', 'lj', 'mol', 'myself', 'pat', 'pat2', 'rainbow', 'snakes', 'tim_reynolds', 'tom', 'train_atkins', 'train_daws', 'train_dotrice', 'train_dreams', 'train_empire', 'train_grace', 'train_kennard', 'train_lescault', 'train_mouse', 'weaver', 'william', 'freeman_2a', 'freeman_3', 'pat4']) in 'parameter_2502' Dropdown component
+export type TortoiseGenerationParams = Seeded & {
+ text: string; // string in 'Prompt' Textbox component
+ speaker: string; // string in 'parameter_2502' Dropdown component
preset: string; // string (Option from: ['ultra_fast', 'fast', 'standard', 'high_quality']) in 'parameter_2507' Dropdown component
- seed: number; // number in 'parameter_2521' Number component
cvvp_amount: number; // number (numeric value between 0.0 and 1.0) in 'CVVP Amount' Slider component
split_prompt: boolean; // boolean in 'Split prompt by lines' Checkbox component
samples: number; // number (numeric value between 4 and 256) in 'Samples' Slider component
@@ -29,11 +35,16 @@ export type TortoiseGenerationParams = {
half: boolean; // boolean in 'parameter_2495' Checkbox component
use_basic_cleaners: boolean; // boolean in 'parameter_2497' Checkbox component
tokenizer: string; // string
+
+ candidates: number;
};
export const initialState: TortoiseGenerationParams = {
+ seed: 0,
+ use_random_seed: true,
+
// autoregressive
- samples: 4,
+ samples: 16,
temperature: 0.8,
length_penalty: 1.0,
repetition_penalty: 2.0,
@@ -48,10 +59,9 @@ export const initialState: TortoiseGenerationParams = {
model: "Default", // string (Option from: ['Default']) in 'parameter_2488' Dropdown component
- prompt: "Howdy!", // string in 'Prompt' Textbox component
- speaker: "random", // string (Option from: ['random', 'angie', 'applejack', 'cond_latent_example', 'daniel', 'deniro', 'emma', 'freeman', 'geralt', 'halle', 'jlaw', 'lj', 'mol', 'myself', 'pat', 'pat2', 'rainbow', 'snakes', 'tim_reynolds', 'tom', 'train_atkins', 'train_daws', 'train_dotrice', 'train_dreams', 'train_empire', 'train_grace', 'train_kennard', 'train_lescault', 'train_mouse', 'weaver', 'william', 'freeman_2a', 'freeman_3', 'pat4']) in 'parameter_2502' Dropdown component
+ text: "Howdy!", // string in 'Prompt' Textbox component
+ speaker: "random", // string in 'parameter_2502' Dropdown component
preset: "ultra_fast", // string (Option from: ['ultra_fast', 'fast', 'standard', 'high_quality']) in 'parameter_2507' Dropdown component
- seed: -1, // number in 'parameter_2521' Number component
cvvp_amount: 0, // number (numeric value between 0.0 and 1.0) in 'CVVP Amount' Slider component
split_prompt: true, // boolean in 'Split prompt by lines' Checkbox component
generation_name: "", // string in 'Generation Name' Textbox component
@@ -61,9 +71,11 @@ export const initialState: TortoiseGenerationParams = {
half: false, // boolean in 'parameter_2495' Checkbox component
use_basic_cleaners: false, // boolean in 'parameter_2497' Checkbox component
tokenizer: "", // string
+
+ candidates: 1,
};
-export const tortoiseGenerationId = "tortoise_generation-tab";
+export const tortoiseGenerationId = "tortoise_generation-tab.v2";
export const sendToBarkVoiceGeneration = (audio?: string) => {
if (!audio) return;
@@ -79,4 +91,52 @@ export const useTortoiseGenerationParams = () =>
useLocalStorage(tortoiseGenerationId, initialState);
export const useTortoiseResult = () =>
- useLocalStorage(tortoiseGenerationId + ".output", null);
+ useLocalStorage(
+ tortoiseGenerationId + ".output",
+ null
+ );
+
+export const getTortoiseGenerationParams = (): TortoiseGenerationParams =>
+ readLocalStorage(tortoiseGenerationId) ?? initialState;
+
+export const useTortoisePage = () => {
+ const [tortoiseGenerationParams, setTortoiseGenerationParams] =
+ useTortoiseGenerationParams();
+
+ const [historyData, setHistoryData] = useHistory("tortoise.v3");
+
+ const consumer = async (params: TortoiseGenerationParams) => {
+ const results = await generateWithTortoise(params);
+ if (params.use_random_seed)
+ setTortoiseGenerationParams((x) => ({ ...x, seed: params.seed }));
+ setHistoryData((x) => [...results, ...x]);
+ return results;
+ };
+
+ const funcs = {
+ favorite: async (_url: string, data?: TortoiseResult) => {
+ const history_bundle_name_data = data?.bundle_name;
+ favorite("", { history_bundle_name_data });
+ },
+ useSeed: useSeedHelper(setTortoiseGenerationParams),
+ // useParameters: (_url: string, data?: TortoiseResult) => {
+ // const params = data?.metadata;
+ // if (!params) return;
+ // setTortoiseGenerationParams({
+ // ...tortoiseGenerationParams,
+ // ...params,
+ // seed: Number(params.seed),
+ // });
+ // },
+ };
+
+ return {
+ tortoiseGenerationParams,
+ setTortoiseGenerationParams,
+ historyData,
+ setHistoryData,
+ consumer,
+ funcs,
+ handleChange: parseFormChange(setTortoiseGenerationParams),
+ };
+};
diff --git a/react-ui/src/tabs/TortoiseResult.tsx b/react-ui/src/tabs/TortoiseResult.tsx
index 5372985f..9b79dbb5 100644
--- a/react-ui/src/tabs/TortoiseResult.tsx
+++ b/react-ui/src/tabs/TortoiseResult.tsx
@@ -6,5 +6,19 @@ export type TortoiseResult = {
image: string;
seed: string;
bundle_name: string;
- metadata: {};
+ metadata: {
+ // Potentially wrong
+ _version: string;
+ _hash_version: string;
+ _type: string;
+ text: string;
+ voice: string;
+ seed: string;
+ name: string;
+ split_prompt: boolean;
+ cvvp_amount: number;
+ preset: string;
+ candidates: number;
+ date: string;
+ };
};
diff --git a/react-ui/src/tabs/VallexParams.tsx b/react-ui/src/tabs/VallexParams.tsx
index b9fcc1b1..3b0df1a8 100644
--- a/react-ui/src/tabs/VallexParams.tsx
+++ b/react-ui/src/tabs/VallexParams.tsx
@@ -1,12 +1,17 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
import { GradioFile } from "../types/GradioFile";
+import { Seeded } from "../types/Seeded";
+import { useHistory } from "../hooks/useHistory";
+import { generateWithVallex } from "../functions/generateWithVallex";
+import { parseFormChange } from "../data/parseFormChange";
const vallex_ID = "vallexParams";
-export type VallexParams = {
+export type VallexParams = Seeded & {
text: string;
prompt: string;
language: string;
@@ -15,6 +20,9 @@ export type VallexParams = {
};
export const initialVallexParams: VallexParams = {
+ seed: 0,
+ use_random_seed: true,
+
text: "",
prompt: "",
language: "Mix",
@@ -51,3 +59,41 @@ export const useVallexParams = () =>
export const useVallexResult = () =>
useLocalStorage(vallex_ID + ".output", null);
+
+export const getVallexParams = (): VallexParams =>
+ readLocalStorage(vallex_ID) ?? initialVallexParams;
+
+export const useVallexPage = () => {
+ const [vallexParams, setVallexParams] = useVallexParams();
+
+ const [historyData, setHistoryData] = useHistory("vallex");
+
+ const consumer = async (params: VallexParams) => {
+ const data = await generateWithVallex(params);
+ if (params.use_random_seed)
+ setVallexParams((x) => ({ ...x, seed: params.seed }));
+ setHistoryData((x) => [data, ...x]);
+ return data;
+ };
+
+ const funcs = {
+ useParameters: (_url: string, data?: VallexResult) => {
+ const params = data?.metadata;
+ if (!params) return;
+ setVallexParams({
+ ...vallexParams,
+ ...params,
+ });
+ },
+ };
+
+ return {
+ vallexParams,
+ setVallexParams,
+ historyData,
+ setHistoryData,
+ consumer,
+ funcs,
+ handleChange: parseFormChange(setVallexParams),
+ };
+};
diff --git a/react-ui/src/tabs/VocosParams.tsx b/react-ui/src/tabs/VocosParams.tsx
index d9bf73c3..03061a26 100644
--- a/react-ui/src/tabs/VocosParams.tsx
+++ b/react-ui/src/tabs/VocosParams.tsx
@@ -1,4 +1,5 @@
import useLocalStorage, {
+ readLocalStorage,
updateLocalStorageWithFunction,
} from "../hooks/useLocalStorage";
import router from "next/router";
@@ -33,3 +34,6 @@ export const useVocosParams = () =>
export const useVocosResults = () =>
useLocalStorage(vocosId + ".output", null);
+
+export const getVocosParams = (): VocosParams =>
+ readLocalStorage(vocosId) ?? initialState;
diff --git a/react-ui/src/types/GradioFile.tsx b/react-ui/src/types/GradioFile.tsx
index a17aa61c..ef6af2c1 100644
--- a/react-ui/src/types/GradioFile.tsx
+++ b/react-ui/src/types/GradioFile.tsx
@@ -2,6 +2,7 @@
export type GradioFile = {
name: string;
data: string;
+ url: string; // appears to be the new 'data'
size?: number;
is_file?: boolean;
orig_name?: string;
diff --git a/react-ui/src/types/HandleChange.tsx b/react-ui/src/types/HandleChange.tsx
index 4bd09a89..d7bbb0ac 100644
--- a/react-ui/src/types/HandleChange.tsx
+++ b/react-ui/src/types/HandleChange.tsx
@@ -3,6 +3,7 @@ import React from "react";
export type HandleChangeEvent =
| React.ChangeEvent
| React.ChangeEvent
- | React.ChangeEvent;
+ | React.ChangeEvent
+ | { target: { name: string; value: string; type?: string } };
export type HandleChange = (event: HandleChangeEvent) => void;
diff --git a/react-ui/src/types/Seeded.tsx b/react-ui/src/types/Seeded.tsx
new file mode 100644
index 00000000..b9e9ae46
--- /dev/null
+++ b/react-ui/src/types/Seeded.tsx
@@ -0,0 +1,4 @@
+export type Seeded = {
+ seed: number;
+ use_random_seed: boolean;
+};
diff --git a/requirements.txt b/requirements.txt
index 1caacec8..cf99b4a1 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,13 +1,15 @@
-safetensors==0.3.1 # Until they fix it
+safetensors
+# safetensors==0.3.1 # Until they fix it
ffmpeg-python # Apache 2.0
-gradio==3.48.0
+# gradio==3.48.0
+gradio>=4.1.0 # MIT License
python-dotenv==1.0.0
soundfile==0.12.1; sys_platform == 'win32' # torchaudio
# sox # torchaudio for linux
-suno-bark @ git+https://github.com/rsxdalv/bark@0d91823ead3d87c317f12d01d325fca9408c669e#egg=suno-bark # MIT License
+suno-bark @ git+https://github.com/rsxdalv/bark@main#egg=suno-bark # MIT License
vocos==0.1.0 # MIT License
-tortoise-tts @ git+https://github.com/rsxdalv/tortoise-tts@e4711433b12bcd1086840649e1830ad5c3fa1a76#egg=tortoise-tts # Apache 2.0
+tortoise-tts @ git+https://github.com/rsxdalv/tortoise-tts@main#egg=tortoise-tts # Apache 2.0
beartype>=0.16.1 # workaround for a bug
-transformers==4.36.1 # cross-compatibility
+# no longer required directly # transformers==4.36.1 # cross-compatibility
iso639-lang==2.2.3
pillow==10.3.0 # for gradio, conda fix
diff --git a/server.py b/server.py
index 3d0c19b7..92a8984c 100644
--- a/server.py
+++ b/server.py
@@ -73,11 +73,43 @@ def load_tabs(list_of_tabs):
run_tab(module_name, function_name, name, requirements)
-def main_ui():
+def main_ui(theme_choice = "Base"):
+ themes = {
+ "Base": gr.themes.Base,
+ "Default": gr.themes.Default,
+ "Monochrome": gr.themes.Monochrome,
+ }
+ theme = themes[theme_choice](
+ # primary_hue="blue",
+ primary_hue="sky",
+ secondary_hue="sky",
+ neutral_hue="neutral",
+ font=[gr.themes.GoogleFont('Inter'), 'ui-sans-serif', 'system-ui', 'sans-serif'],
+ ).set(
+ embed_radius='*radius_sm',
+ block_label_radius='*radius_sm',
+ block_label_right_radius='*radius_sm',
+ block_radius='*radius_sm',
+ block_title_radius='*radius_sm',
+ container_radius='*radius_sm',
+ checkbox_border_radius='*radius_sm',
+ input_radius='*radius_sm',
+ table_radius='*radius_sm',
+ button_large_radius='*radius_sm',
+ button_small_radius='*radius_sm',
+
+
+ button_primary_background_fill_hover='*primary_300',
+ button_primary_background_fill_hover_dark='*primary_600',
+ button_secondary_background_fill_hover='*secondary_200',
+ button_secondary_background_fill_hover_dark='*secondary_600',
+ )
+
with gr.Blocks(
css=full_css,
title="TTS Generation WebUI",
analytics_enabled=False, # it broke too many times
+ theme=theme,
) as blocks:
gr.Markdown(
"""
@@ -129,7 +161,7 @@ def all_tabs():
]
load_tabs(tts_tabs)
- handle_extension_class("text-to-speech")
+ handle_extension_class("text-to-speech", config)
with gr.Tab("Audio/Music Generation"), gr.Tabs():
audio_music_generation_tabs = [
(
@@ -153,7 +185,7 @@ def all_tabs():
]
load_tabs(audio_music_generation_tabs)
- handle_extension_class("audio-music-generation")
+ handle_extension_class("audio-music-generation", config)
with gr.Tab("Audio Conversion"), gr.Tabs():
audio_conversion_tabs = [
(
@@ -173,17 +205,20 @@ def all_tabs():
]
load_tabs(audio_conversion_tabs)
- handle_extension_class("audio-conversion")
+ handle_extension_class("audio-conversion", config)
with gr.Tab("Outputs"), gr.Tabs():
from src.history_tab.main import history_tab
collections_directories_atom.render()
- history_tab()
- history_tab(directory="favorites")
- history_tab(
- directory="outputs",
- show_collections=True,
- )
+ try:
+ history_tab()
+ history_tab(directory="favorites")
+ history_tab(
+ directory="outputs",
+ show_collections=True,
+ )
+ except Exception as e:
+ generic_error_tab_advanced(e, name="History", requirements=None)
outputs_tabs = [
# voices
@@ -191,13 +226,13 @@ def all_tabs():
]
load_tabs(outputs_tabs)
- handle_extension_class("outputs")
+ handle_extension_class("outputs", config)
with gr.Tab("Tools"), gr.Tabs():
tools_tabs = []
load_tabs(tools_tabs)
- handle_extension_class("tools")
+ handle_extension_class("tools", config)
with gr.Tab("Settings"), gr.Tabs():
from src.settings_tab_gradio import settings_tab_gradio
@@ -217,7 +252,7 @@ def all_tabs():
extension_list_tab()
- handle_extension_class("settings")
+ handle_extension_class("settings", config)
def start_gradio_server():
@@ -244,6 +279,12 @@ def print_pretty_options(options):
gradio_interface_options["auth"].split(":")
)
print("Gradio server authentication enabled")
+ # delete show_tips option
+ if "show_tips" in gradio_interface_options:
+ del gradio_interface_options["show_tips"]
+ # TypeError: Blocks.launch() got an unexpected keyword argument 'file_directories'
+ if "file_directories" in gradio_interface_options:
+ del gradio_interface_options["file_directories"]
print_pretty_options(gradio_interface_options)
demo = main_ui()
@@ -257,8 +298,8 @@ def print_pretty_options(options):
)
demo.queue(
- concurrency_count=gradio_interface_options.get("concurrency_count", 5),
- ).launch(**gradio_interface_options)
+ # concurrency_count=gradio_interface_options.get("concurrency_count", 5),
+ ).launch(**gradio_interface_options, allowed_paths=["."])
if __name__ == "__main__":
diff --git a/src/bark/clone/tab_voice_clone.py b/src/bark/clone/tab_voice_clone.py
index 6e525d4b..8a09a5df 100644
--- a/src/bark/clone/tab_voice_clone.py
+++ b/src/bark/clone/tab_voice_clone.py
@@ -183,7 +183,7 @@ def tab_voice_clone():
file_input = gr.Audio(
label="Input Audio",
type="filepath",
- source="upload",
+ sources="upload",
interactive=True,
)
with gr.Row():
@@ -199,7 +199,7 @@ def clear_models():
hubert_model = None
tokenizer = None
torch.cuda.empty_cache()
- return gr.Button.update(
+ return gr.Button(
value="Models cleared",
)
diff --git a/src/bark/generate_choice_string.py b/src/bark/generate_choice_string.py
index 7725729f..8207a14d 100644
--- a/src/bark/generate_choice_string.py
+++ b/src/bark/generate_choice_string.py
@@ -7,4 +7,4 @@
def generate_choice_string(use_v2: bool, language: str, speaker_id: Union[int, str]):
history_prompt = create_voice_string(language, speaker_id, use_v2)
gender = get_speaker_gender(history_prompt)
- return gr.Markdown.update(value=f"Chosen voice: {history_prompt}, Gender: {gender}")
+ return gr.Markdown(value=f"Chosen voice: {history_prompt}, Gender: {gender}")
diff --git a/src/bark/generation_tab_bark.py b/src/bark/generation_tab_bark.py
index 9553481f..33509034 100644
--- a/src/bark/generation_tab_bark.py
+++ b/src/bark/generation_tab_bark.py
@@ -33,6 +33,7 @@
)
from src.magnet.utils import Timer
from src.utils.prompt_to_title import prompt_to_title
+from src.utils.randomize_seed import randomize_seed
FREEZE_SEMANTIC = False
@@ -192,7 +193,7 @@ def gen_helper(
)
yield yield_generation(
- audio=gr.Audio.update(
+ audio=gr.Audio(
value=filename,
label=f"Generated audio fragment... `{prompt_piece}`",
),
@@ -217,7 +218,7 @@ def gen_helper(
)
yield yield_generation(
- audio=gr.Audio.update(value=filename, label="Generated audio"),
+ audio=gr.Audio(value=filename, label="Generated audio"),
npz=filename_npz,
json_text=metadata,
history_bundle_name_data=os.path.dirname(filename),
@@ -255,7 +256,7 @@ def bark_ui():
def unload_models():
model_manager.unload_models()
- return gr.Button.update(value="Unloaded")
+ return gr.Button(value="Unloaded")
unload_models_button = gr.Button("Unload models")
unload_models_button.click(
@@ -272,7 +273,7 @@ def unload_models():
# Show the language and speakerId radios only when useHistory is checked
history_setting.change(
- fn=lambda choice: gr.Column.update(visible=(choice == HistorySettings.VOICE)),
+ fn=lambda choice: gr.Column(visible=(choice == HistorySettings.VOICE)),
inputs=[history_setting],
outputs=[column], # type: ignore
)
@@ -291,9 +292,9 @@ def unload_models():
history_setting.change(
fn=lambda choice: [
- gr.Dropdown.update(visible=(choice == HistorySettings.NPZ_FILE)),
- gr.Button.update(visible=(choice == HistorySettings.NPZ_FILE)),
- gr.Button.update(visible=(choice == HistorySettings.NPZ_FILE)),
+ gr.Dropdown(visible=(choice == HistorySettings.NPZ_FILE)),
+ gr.Button(visible=(choice == HistorySettings.NPZ_FILE)),
+ gr.Button(visible=(choice == HistorySettings.NPZ_FILE)),
],
inputs=[history_setting],
outputs=[
@@ -305,9 +306,9 @@ def unload_models():
history_setting.change(
fn=lambda choice: [
- gr.Dropdown.update(visible=(choice == HistorySettings.NPZ_FILE)),
- gr.Button.update(visible=(choice == HistorySettings.NPZ_FILE)),
- gr.Button.update(visible=(choice == HistorySettings.NPZ_FILE)),
+ gr.Dropdown(visible=(choice == HistorySettings.NPZ_FILE)),
+ gr.Button(visible=(choice == HistorySettings.NPZ_FILE)),
+ gr.Button(visible=(choice == HistorySettings.NPZ_FILE)),
],
inputs=[history_setting],
outputs=[
@@ -394,13 +395,6 @@ def unload_models():
label="Randomize seed", value=True
)
-
- def randomize_seed(seed, randomize_seed):
- if randomize_seed:
- return np.random.randint(0, 2**32 - 1, dtype=np.uint32)
- else:
- return int(seed)
-
burn_in_prompt = gr.Textbox(
label="Burn In Prompt (Optional)", lines=3, placeholder="Enter text here..."
)
@@ -460,9 +454,9 @@ def randomize_seed(seed, randomize_seed):
# yield yield_generation(
# audio=None,
- # save_button=gr.Button.update(value="Save"),
- # continue_button=gr.Button.update(),
- # buttons_row=gr.Row.update(visible=False),
+ # save_button=gr.Button(value="Save"),
+ # continue_button=gr.Button(),
+ # buttons_row=gr.Row(visible=False),
# npz=None,
# json_text=None,
# history_bundle_name_data=None,
@@ -527,7 +521,7 @@ def old_generation_dropdown_ui(label):
)
reload_old_generation_dropdown.click(
- fn=lambda: gr.Dropdown.update(choices=get_npz_files()), # type: ignore
+ fn=lambda: gr.Dropdown(choices=get_npz_files()), # type: ignore
outputs=old_generation_dropdown,
api_name=f"reload_old_generation_dropdown{ '' if label == 'Audio Voice' else '_semantic'}",
)
@@ -573,6 +567,6 @@ def setup_bark_voice_prompt_ui():
def insert_npz_file(npz_filename):
return [
- gr.Dropdown.update(value=npz_filename),
- gr.Radio.update(value=HistorySettings.NPZ_FILE),
+ gr.Dropdown(value=npz_filename),
+ gr.Radio(value=HistorySettings.NPZ_FILE),
]
diff --git a/src/bark/settings_tab_bark.py b/src/bark/settings_tab_bark.py
index a3395353..99a950dd 100644
--- a/src/bark/settings_tab_bark.py
+++ b/src/bark/settings_tab_bark.py
@@ -26,7 +26,7 @@ def settings_tab_bark() -> None:
)
with gr.Row(variant="panel"):
- gr.Markdown("### Coarse-to-fine inference:", scale=2)
+ gr.Markdown("### Coarse-to-fine inference:")
coarse_use_gpu = gr.Checkbox(
label="Use GPU",
value=config["model"]["coarse_use_gpu"],
@@ -151,7 +151,7 @@ def save_environment_variables(
def sync_ui():
def checkbox_update_helper(key: str):
- return gr.Checkbox.update(value=config["model"][key])
+ return gr.Checkbox(value=config["model"][key])
return [
checkbox_update_helper("text_use_gpu"),
@@ -161,13 +161,13 @@ def checkbox_update_helper(key: str):
checkbox_update_helper("fine_use_gpu"),
checkbox_update_helper("fine_use_small"),
checkbox_update_helper("codec_use_gpu"),
- gr.Checkbox.update(value=config["load_models_on_startup"]),
+ gr.Checkbox(value=config["load_models_on_startup"]),
]
settings_tab.select(fn=sync_ui, outputs=inputs, api_name="get_config_bark")
def set_to_reload():
- return gr.Button.update(value="Loading...", interactive=False)
+ return gr.Button(value="Loading...", interactive=False)
with gr.Column():
gr.Markdown(
@@ -226,7 +226,7 @@ def load_models(
)
try:
model_manager.reload_models(config)
- return gr.Button.update(value="Reload models", interactive=True)
+ return gr.Button(value="Reload models", interactive=True)
except Exception as e:
print(e)
- return gr.Button.update(value="Failed to load models", interactive=True)
+ return gr.Button(value="Failed to load models", interactive=True)
diff --git a/src/bark/setup_seed_ui_bark.py b/src/bark/setup_seed_ui_bark.py
index 87be95ff..8fa1151e 100644
--- a/src/bark/setup_seed_ui_bark.py
+++ b/src/bark/setup_seed_ui_bark.py
@@ -11,7 +11,7 @@ def setup_seed_ui_bark():
)
set_random_seed_button.click(
- fn=lambda: gr.Textbox.update(value="-1"), outputs=[seed_input]
+ fn=lambda: gr.Textbox(value="-1"), outputs=[seed_input]
)
set_old_seed_button = gr.Button(
diff --git a/src/config/load_config.py b/src/config/load_config.py
index 3aa12a36..b131adc9 100644
--- a/src/config/load_config.py
+++ b/src/config/load_config.py
@@ -25,7 +25,7 @@
"show_error": False,
"server_name": "0.0.0.0",
"server_port": 7770,
- "show_tips": False,
+ # "show_tips": False, # DEPRECATED
"height": 500,
"width": "100%",
"favicon_path": None,
@@ -39,6 +39,9 @@
"_frontend": True,
},
"load_models_on_startup": False,
+ "extensions": {
+ "disabled": [],
+ },
}
diff --git a/src/css/material_symbols_css.py b/src/css/material_symbols_css.py
index 96b4d84a..8a47f8a1 100644
--- a/src/css/material_symbols_css.py
+++ b/src/css/material_symbols_css.py
@@ -9,7 +9,6 @@
line-height: 1 !important;
letter-spacing: normal;
text-transform: none;
- display: inline-block;
white-space: nowrap;
word-wrap: normal;
direction: ltr;
@@ -17,3 +16,4 @@
-webkit-font-smoothing: antialiased;
}
"""
+# display: inline-block;
diff --git a/src/extensions_loader/interface_extensions.py b/src/extensions_loader/interface_extensions.py
index 6eda113a..65cd889d 100644
--- a/src/extensions_loader/interface_extensions.py
+++ b/src/extensions_loader/interface_extensions.py
@@ -6,7 +6,7 @@
import gradio as gr
-from src.utils.pip_install import pip_install_wrapper
+from src.utils.pip_install import pip_install_wrapper, pip_uninstall_wrapper
from src.utils.generic_error_tab_advanced import generic_error_tab_advanced
@@ -15,7 +15,14 @@ def check_if_package_installed(package_name):
return spec is not None
+# disabled_extensions = ["extension_xtts_rvc_ui"]
+disabled_extensions = []
+
+
def _handle_package(package_name, title_name, requirements):
+ if package_name in disabled_extensions:
+ print(f"Skipping disabled {title_name} Extension...")
+ return
if check_if_package_installed(package_name):
print(f"Loading {title_name} Extension...")
start_time = time.time()
@@ -71,7 +78,7 @@ def get_extension_list_json():
extension_list_json = get_extension_list_json()
-def handle_extension_class(extension_class):
+def handle_extension_class(extension_class, config):
for x in extension_list_json:
if (
x["extension_type"] == "interface"
@@ -93,6 +100,49 @@ def extension_list_tab():
)
gr.Markdown(table_string)
+ external_extension_list = [
+ x for x in extension_list_json if "builtin" not in x["package_name"]
+ ]
+
+ with gr.Row():
+ with gr.Column():
+ gr.Markdown("Install/Uninstall Extensions")
+
+ install_dropdown = gr.Dropdown(
+ label="Select Extension to Install",
+ choices=[x["package_name"] for x in external_extension_list],
+ )
+
+ install_button = gr.Button("Install extension")
+
+ def install_extension(package_name):
+ yield from pip_install_wrapper(package_name, package_name)()
+
+ install_button.click(
+ fn=install_extension,
+ inputs=[install_dropdown],
+ outputs=[gr.HTML()],
+ api_name="install_extension",
+ )
+
+ with gr.Column():
+ gr.Markdown("Uninstall Extensions")
+ uninstall_dropdown = gr.Dropdown(
+ label="Select Extension to Uninstall",
+ choices=[x["package_name"] for x in external_extension_list],
+ )
+ uninstall_button = gr.Button("Uninstall extension")
+
+ def uninstall_extension(package_name):
+ yield from pip_uninstall_wrapper(package_name, package_name)()
+
+ uninstall_button.click(
+ fn=uninstall_extension,
+ inputs=[uninstall_dropdown],
+ outputs=[gr.HTML()],
+ api_name="uninstall_extension",
+ )
+
if __name__ == "__main__":
with gr.Blocks() as demo:
diff --git a/src/history_tab/edit_metadata_ui.py b/src/history_tab/edit_metadata_ui.py
index 53f303ea..3f399a51 100644
--- a/src/history_tab/edit_metadata_ui.py
+++ b/src/history_tab/edit_metadata_ui.py
@@ -17,8 +17,8 @@ def save_metadata(metadata_input, voice_file_name):
"semantic_prompt": full_generation["semantic_prompt"],
}, metadata=metadata)
return {
- metadata_display: gr.JSON.update(value=metadata),
- save_metadata_button: gr.Button.update(value="Saved"),
+ metadata_display: gr.JSON(value=metadata),
+ save_metadata_button: gr.Button(value="Saved"),
}
save_metadata_button.click(
@@ -28,7 +28,7 @@ def save_metadata(metadata_input, voice_file_name):
)
metadata_input.change(
- fn=lambda: save_metadata_button.update(value="Save Metadata"),
+ fn=lambda: gr.Button(value="Save Metadata"),
outputs=[save_metadata_button]
)
diff --git a/src/history_tab/main.py b/src/history_tab/main.py
index 0bb07efe..c57250fb 100644
--- a/src/history_tab/main.py
+++ b/src/history_tab/main.py
@@ -33,16 +33,16 @@ def _select_audio(table, evt: gr.SelectData):
def clear_audio():
return [
- gr.Audio.update(value=None),
- gr.Image.update(value=None),
- gr.JSON.update(value=None),
- gr.Button.update(visible=False),
+ gr.Audio(value=None),
+ gr.Image(value=None),
+ gr.JSON(value=None),
+ gr.Button(visible=False),
]
def save_to_voices_cb(npz_filename: str):
shutil.copy(npz_filename, "voices/")
- return gr.Button.update(value="Saved")
+ return gr.Button(value="Saved")
def history_tab(directory="outputs", show_collections=False):
@@ -61,7 +61,7 @@ def history_content(directory, history_tab, show_collections):
visible=show_collections,
)
collections_directories_atom.change(
- fn=lambda x: gr.Dropdown.update(choices=x),
+ fn=lambda x: gr.Dropdown(choices=x),
inputs=[collections_directories_atom],
outputs=[directory_dropdown],
)
@@ -96,7 +96,6 @@ def history_content(directory, history_tab, show_collections):
type="array",
interactive=False,
col_count=len(datatypes),
- max_cols=len(datatypes),
datatype=datatypes,
headers=headers,
height=800,
@@ -148,22 +147,22 @@ def history_content(directory, history_tab, show_collections):
def _select_audio_history(filename: str, json_text):
return {
- history_bundle_name: gr.Textbox.update(value=os.path.dirname(filename)),
+ history_bundle_name: gr.Textbox(value=os.path.dirname(filename)),
history_bundle_name_data: os.path.dirname(filename),
- history_audio: gr.Audio.update(value=filename, label=filename),
+ history_audio: gr.Audio(value=filename, label=filename),
history_image: (
- gr.Image.update(value=filename.replace(".wav", ".png"))
+ gr.Image(value=filename.replace(".wav", ".png"))
if os.path.exists(filename.replace(".wav", ".png"))
- else gr.Image.update(value=None)
+ else gr.Image(value=None)
),
- history_json: gr.JSON.update(value=json_text),
- history_npz: gr.Textbox.update(value=filename.replace(".wav", ".npz")),
- delete_from_history: gr.Button.update(visible=True),
- save_to_favorites_history: gr.Button.update(
+ history_json: gr.JSON(value=json_text),
+ history_npz: gr.Textbox(value=filename.replace(".wav", ".npz")),
+ delete_from_history: gr.Button(visible=True),
+ save_to_favorites_history: gr.Button(
visible=directory != "favorites", value="Save to favorites"
),
- save_to_voices: gr.Button.update(visible=True, value="Save to voices"),
- open_folder_button: gr.Button.update(visible=True),
+ save_to_voices: gr.Button(visible=True, value="Save to voices"),
+ open_folder_button: gr.Button(visible=True),
}
def select_audio_history(table, evt: gr.SelectData):
@@ -192,19 +191,8 @@ def select_audio_history2(_list, evt: gr.SelectData, table):
preprocess=False,
)
- # history_list_as_gallery.select(
- # fn=select_audio_history2,
- # inputs=[history_list_as_gallery, history_list],
- # outputs=outputs,
- # preprocess=False,
- # )
-
def update_history_tab(directory: str):
- # return [
- # gr.Dataframe.update(value=get_wav_files(directory)),
- # gr.Gallery.update(value=get_wav_files_img(directory)),
- # ]
- return gr.Dataframe.update(value=get_wav_files(directory))
+ return gr.Dataframe(value=get_wav_files(directory))
delete_from_history.click(
fn=clear_audio,
@@ -218,7 +206,7 @@ def update_history_tab(directory: str):
)
# API ONLY
gr.Button(
- label="Delete (API ONLY)",
+ value="Delete (API ONLY)",
visible=False,
).click(
fn=delete_generation,
@@ -240,7 +228,7 @@ def update_history_tab(directory: str):
)
reload_button.click(
- fn=lambda x: gr.Dataframe.update(value=get_wav_files(x)),
+ fn=lambda x: gr.Dataframe(value=get_wav_files(x)),
inputs=[directory_dropdown],
outputs=[history_list],
api_name=f"{'collections' if show_collections else directory}_refresh_history",
@@ -267,7 +255,7 @@ def save_to_collection_ui(
)
directories_state.change(
- fn=lambda x: gr.Dropdown.update(choices=x),
+ fn=lambda x: gr.Dropdown(choices=x),
inputs=[directories_state],
outputs=[move_to_collection],
)
@@ -280,13 +268,13 @@ def create_collection(new_collection_name):
os.makedirs(os.path.join("collections", new_collection_name))
return [
get_collections(),
- gr.Button.update(value="Created"),
+ gr.Button(value="Created"),
]
create_collection_button = gr.Button(value="Create collection")
new_collection_name.change(
- fn=lambda: gr.Button.update(value="Create collection"),
+ fn=lambda: gr.Button(value="Create collection"),
outputs=[create_collection_button],
)
create_collection_button.click(
diff --git a/src/history_tab/save_to_favorites.py b/src/history_tab/save_to_favorites.py
index c37269bd..db695cff 100644
--- a/src/history_tab/save_to_favorites.py
+++ b/src/history_tab/save_to_favorites.py
@@ -6,9 +6,9 @@
def save_to_favorites(directory: str):
shutil.copytree(directory, os.path.join("favorites", os.path.basename(directory)))
- return gr.Button.update(value="Saved")
+ return gr.Button(value="Saved")
def save_to_collection(directory: str, collection: str):
shutil.copytree(directory, os.path.join(collection, os.path.basename(directory)))
- return gr.Dropdown.update(value="Saved")
+ return gr.Dropdown(value="Saved")
diff --git a/src/history_tab/voices_tab.py b/src/history_tab/voices_tab.py
index 89ef0bda..7e37c302 100644
--- a/src/history_tab/voices_tab.py
+++ b/src/history_tab/voices_tab.py
@@ -15,7 +15,7 @@
def update_voices_tab():
- return gr.List.update(value=get_npz_files_voices())
+ return gr.List(value=get_npz_files_voices())
def voices_tab(directory="voices"):
@@ -52,7 +52,6 @@ def voices_tab(directory="voices"):
interactive=False,
datatype=datatypes,
col_count=len(datatypes),
- max_cols=len(datatypes),
headers=headers,
height=800,
# elem_classes="file-list"
@@ -86,7 +85,7 @@ def voices_tab(directory="voices"):
def delete_voice(voice_file_name):
os.remove(voice_file_name)
return {
- delete_voice_button: gr.Button.update(value="Deleted"),
+ delete_voice_button: gr.Button(value="Deleted"),
voices_list: update_voices_tab(),
}
@@ -96,9 +95,9 @@ def rename_voice(voice_file_name_in, new_voice_file_name):
if os.path.exists(png_file):
shutil.move(png_file, new_voice_file_name.replace(".npz", ".png"))
return {
- rename_voice_button: gr.Button.update(value="Renamed"),
+ rename_voice_button: gr.Button(value="Renamed"),
voices_list: update_voices_tab(),
- voice_file_name: gr.Textbox.update(value=new_voice_file_name),
+ voice_file_name: gr.Textbox(value=new_voice_file_name),
}
def crop_voice(voice_file_name, audio_in):
@@ -176,18 +175,18 @@ def select_filename(filename_npz):
resolved_photo = None
return {
- voice_file_name: gr.Textbox.update(value=filename_npz),
- new_voice_file_name: gr.Textbox.update(value=filename_npz),
- delete_voice_button: gr.Button.update(value="Delete"),
- rename_voice_button: gr.Button.update(value="Rename"),
- audio: gr.Audio.update(value=get_audio_from_full_generation(full_generation)), # type: ignore
- metadata: gr.JSON.update(value=full_generation.get("metadata", {})),
- metadata_input: gr.Textbox.update(
+ voice_file_name: gr.Textbox(value=filename_npz),
+ new_voice_file_name: gr.Textbox(value=filename_npz),
+ delete_voice_button: gr.Button(value="Delete"),
+ rename_voice_button: gr.Button(value="Rename"),
+ audio: gr.Audio(value=get_audio_from_full_generation(full_generation)), # type: ignore
+ metadata: gr.JSON(value=full_generation.get("metadata", {})),
+ metadata_input: gr.Textbox(
value=json.dumps(full_generation.get("metadata", {}), indent=2)
),
- photo: gr.Image.update(value=resolved_photo),
- voice_hash: gr.Textbox.update(value=history_to_hash(full_generation)), # type: ignore
- file_list: gr.Files.update(
+ photo: gr.Image(value=resolved_photo),
+ voice_hash: gr.Textbox(value=history_to_hash(full_generation)), # type: ignore
+ file_list: gr.Files(
value=get_file_list(filename_npz, resolved_photo),
label="Files",
),
diff --git a/src/magnet/magnet_tab.py b/src/magnet/magnet_tab.py
index 51a08ecf..571475a3 100644
--- a/src/magnet/magnet_tab.py
+++ b/src/magnet/magnet_tab.py
@@ -282,7 +282,7 @@ def generation_tab_magnet():
value=initial_params["model"],
)
gr_reload_button().click(
- fn=lambda: gr.Radio.update(choices=get_models()), # type: ignore
+ fn=lambda: gr.Radio(choices=get_models()), # type: ignore
outputs=[model],
api_name="magnet_get_models",
)
@@ -357,7 +357,7 @@ def generation_tab_magnet():
# )
# set_random_seed_button.click(
- # fn=lambda: gr.Number.update(value=-1), outputs=[seed_input]
+ # fn=lambda: gr.Number(value=-1), outputs=[seed_input]
# )
# set_old_seed_button = gr.Button(
@@ -366,7 +366,7 @@ def generation_tab_magnet():
# def link_seed_cache(seed_cache):
# set_old_seed_button.click(
- # fn=lambda x: gr.Number.update(value=x),
+ # fn=lambda x: gr.Number(value=x),
# inputs=seed_cache,
# outputs=seed_input,
# )
diff --git a/src/maha_tts/maha_tts_tab.py b/src/maha_tts/maha_tts_tab.py
index 2fbe5ab5..cdabd0cd 100644
--- a/src/maha_tts/maha_tts_tab.py
+++ b/src/maha_tts/maha_tts_tab.py
@@ -197,12 +197,12 @@ def maha_tts_ui():
)
model_language.change(
fn=lambda choice: choice == "Smolie-en"
- and gr.Radio.update(
+ and gr.Radio(
value="english",
visible=False,
interactive=False,
)
- or gr.Radio.update(
+ or gr.Radio(
interactive=True,
visible=True,
),
@@ -210,7 +210,7 @@ def maha_tts_ui():
outputs=[maha_tts_language],
)
- with gr.Box():
+ with gr.Column():
gr.Markdown("Speaker Name")
with gr.Row():
voices = get_voice_list()
@@ -223,7 +223,7 @@ def maha_tts_ui():
)
gr_open_button_simple("voices-tortoise", api_name="maha_tts_open_voices")
gr_reload_button().click(
- fn=lambda: gr.Dropdown.update(choices=get_voice_list()), # type: ignore
+ fn=lambda: gr.Dropdown(choices=get_voice_list()), # type: ignore
outputs=[speaker_name],
api_name="maha_tts_refresh_voices",
)
diff --git a/src/musicgen/musicgen_tab.py b/src/musicgen/musicgen_tab.py
index 7ee92b89..778f0520 100644
--- a/src/musicgen/musicgen_tab.py
+++ b/src/musicgen/musicgen_tab.py
@@ -278,7 +278,7 @@ def generation_tab_musicgen():
value="facebook/musicgen-small",
)
melody = gr.Audio(
- source="upload",
+ sources="upload",
type="numpy",
label="Melody (optional)",
elem_classes="tts-audio",
@@ -334,7 +334,7 @@ def generation_tab_musicgen():
interactive=False,
elem_classes="tts-audio",
)
- image = gr.Image(label="Waveform", shape=(None, 100), elem_classes="tts-image") # type: ignore
+ image = gr.Image(label="Waveform", elem_classes="tts-image", visible=False) # type: ignore
with gr.Row():
history_bundle_name_data = gr.Textbox(visible=False)
save_button = gr.Button("Save to favorites", visible=True)
@@ -395,7 +395,7 @@ def update_json(
)
set_old_seed_button.click(
- fn=lambda x: gr.Number.update(value=x),
+ fn=lambda x: gr.Number(value=x),
inputs=seed_cache,
outputs=seed,
)
diff --git a/src/musicgen/setup_seed_ui_musicgen.py b/src/musicgen/setup_seed_ui_musicgen.py
index 2a5ea2b8..723f4898 100644
--- a/src/musicgen/setup_seed_ui_musicgen.py
+++ b/src/musicgen/setup_seed_ui_musicgen.py
@@ -10,7 +10,7 @@ def setup_seed_ui_musicgen():
)
set_random_seed_button.click(
- fn=lambda: gr.Number.update(value=-1), outputs=[seed_input]
+ fn=lambda: gr.Number(value=-1), outputs=[seed_input]
)
set_old_seed_button = gr.Button(
@@ -19,7 +19,7 @@ def setup_seed_ui_musicgen():
def link_seed_cache(seed_cache):
set_old_seed_button.click(
- fn=lambda x: gr.Number.update(value=x),
+ fn=lambda x: gr.Number(value=x),
inputs=seed_cache,
outputs=seed_input,
)
diff --git a/src/rvc_tab/rvc_tab.py b/src/rvc_tab/rvc_tab.py
index 08f463b2..91b34d47 100644
--- a/src/rvc_tab/rvc_tab.py
+++ b/src/rvc_tab/rvc_tab.py
@@ -124,7 +124,7 @@ def get_rvc_index_list():
def rvc_ui_model_or_index_path_ui(label: str):
get_list_fn = get_rvc_model_list if label == "Model" else get_rvc_index_list
- with gr.Box():
+ with gr.Column():
gr.Markdown(f"{label}")
with gr.Row():
file_path_dropdown = gr.Dropdown(
@@ -137,9 +137,7 @@ def rvc_ui_model_or_index_path_ui(label: str):
RVC_LOCAL_MODELS_DIR, api_name=f"rvc_{label.lower()}_open"
)
gr_reload_button().click(
- lambda: file_path_dropdown.update(
- choices=get_list_fn(), # type: ignore
- ),
+ fn=lambda: gr.Dropdown(choices=get_list_fn()),
outputs=[file_path_dropdown],
api_name=f"rvc_{label.lower()}_reload",
)
diff --git a/src/seamlessM4T/seamless_tab.py b/src/seamlessM4T/seamless_tab.py
index 8a47c759..b0e29308 100644
--- a/src/seamlessM4T/seamless_tab.py
+++ b/src/seamlessM4T/seamless_tab.py
@@ -91,7 +91,7 @@ def seamless_ui():
button = gr.Button("Translate Text to Speech")
with gr.Tab(label="Audio to Speech"):
input_audio = gr.Audio(
- source="upload",
+ sources="upload",
type="numpy",
label="Input Audio",
elem_classes="tts-audio",
diff --git a/src/settings_tab_gradio.py b/src/settings_tab_gradio.py
index bb52a60b..f083bb71 100644
--- a/src/settings_tab_gradio.py
+++ b/src/settings_tab_gradio.py
@@ -166,5 +166,5 @@ def settings_tab_gradio(
)
reload_config_and_restart_ui_button.click(
- fn=reload_config_and_restart_ui, _js="window.location.reload();"
+ fn=reload_config_and_restart_ui, js="window.location.reload();"
)
diff --git a/src/styletts2/styletts2_tab.py b/src/styletts2/styletts2_tab.py
index a91e311b..266643a3 100644
--- a/src/styletts2/styletts2_tab.py
+++ b/src/styletts2/styletts2_tab.py
@@ -99,10 +99,10 @@ def style_tts2_ui():
reset_params_button = gr.Button("Reset params")
reset_params_button.click(
fn=lambda: [
- alpha.update(0.3),
- beta.update(0.7),
- diffusion_steps.update(5),
- embedding_scale.update(1.0),
+ gr.Slider(value=0.3),
+ gr.Slider(value=0.7),
+ gr.Slider(value=5),
+ gr.Slider(value=1.0),
],
outputs=[
alpha,
diff --git a/src/tortoise/TortoiseOutputRow.py b/src/tortoise/TortoiseOutputRow.py
deleted file mode 100644
index b15b5f2d..00000000
--- a/src/tortoise/TortoiseOutputRow.py
+++ /dev/null
@@ -1,61 +0,0 @@
-import gradio as gr
-
-
-class TortoiseOutputRow:
- def __init__(
- self,
- audio: gr.Audio,
- image: gr.Image,
- save_button: gr.Button,
- seed: gr.Textbox,
- bundle_name: gr.Textbox,
- params: gr.JSON,
- ):
- self.audio: gr.Audio = audio
- self.image: gr.Image = image
- self.save_button: gr.Button = save_button
- self.seed: gr.Textbox = seed
- self.bundle_name: gr.Textbox = bundle_name
- self.params: gr.JSON = params
-
- def to_list(self):
- return [
- self.audio,
- self.image,
- self.save_button,
- self.seed,
- self.bundle_name,
- self.params,
- ]
-
- @staticmethod
- def from_list(components):
- return TortoiseOutputRow(
- audio=components[0],
- image=components[1],
- save_button=components[2],
- seed=components[3],
- bundle_name=components[4],
- params=components[5],
- )
-
- # def __iter__(self):
- # return iter(self.to_list())
-
-
-class TortoiseOutputUpdate:
- def __init__(
- self,
- audio,
- image,
- save_button,
- seed,
- bundle_name,
- params,
- ):
- self.audio = audio
- self.image = image
- self.save_button = save_button
- self.seed = seed
- self.bundle_name = bundle_name
- self.params = params
diff --git a/src/tortoise/TortoiseParameters.py b/src/tortoise/TortoiseParameters.py
index 4c2ccb3b..11dd9338 100644
--- a/src/tortoise/TortoiseParameters.py
+++ b/src/tortoise/TortoiseParameters.py
@@ -92,7 +92,7 @@ def __init__(
text: gr.Textbox,
voice: gr.Dropdown,
preset: gr.Dropdown,
- seed: gr.Number,
+ seed: gr.Textbox,
cvvp_amount: gr.Slider,
split_prompt: gr.Checkbox,
num_autoregressive_samples: gr.Slider,
@@ -207,7 +207,7 @@ def next_idx():
],
value="ultra_fast",
),
- seed=gr.Number(label="Seed", value=None, step=1),
+ seed=gr.Textbox(label="Seed", value=None),
cvvp_amount=gr.Slider(
label="CVVP Amount", value=0.0, minimum=0.0, maximum=1.0, step=0.1
),
diff --git a/src/tortoise/create_tortoise_output_row_ui.py b/src/tortoise/create_tortoise_output_row_ui.py
deleted file mode 100644
index fa57f13d..00000000
--- a/src/tortoise/create_tortoise_output_row_ui.py
+++ /dev/null
@@ -1,36 +0,0 @@
-from src.history_tab.save_to_favorites import save_to_favorites
-from src.tortoise.TortoiseOutputRow import TortoiseOutputRow
-
-
-import gradio as gr
-
-
-def create_tortoise_output_row_ui(index):
- with gr.Column(visible=index == 0) as col:
- audio = gr.Audio(
- type="filepath", label="Generated audio", elem_classes="tts-audio"
- )
- image = gr.Image(label="Waveform", shape=(None, 100), elem_classes="tts-image") # type: ignore
- with gr.Row():
- save_button = gr.Button("Save to favorites", visible=False)
- seed = gr.Textbox(
- visible=False,
- )
- bundle_name = gr.Textbox(
- visible=False,
- )
- params = gr.JSON(
- visible=False,
- )
-
- save_button.click(
- fn=save_to_favorites,
- inputs=[bundle_name],
- outputs=[save_button],
- )
-
- return (
- TortoiseOutputRow(audio, image, save_button, seed, bundle_name, params).to_list(),
- col,
- seed,
- )
diff --git a/src/tortoise/gen_tortoise.py b/src/tortoise/gen_tortoise.py
index 644a60af..a9e0a369 100644
--- a/src/tortoise/gen_tortoise.py
+++ b/src/tortoise/gen_tortoise.py
@@ -1,13 +1,12 @@
import os
-from src.bark.split_text_functions import split_by_lines
import numpy as np
+import gradio as gr
+from src.bark.split_text_functions import split_by_lines
from src.utils.create_base_filename import create_base_filename
from src.utils.date import get_date_string
from src.utils.save_waveform_plot import middleware_save_waveform_plot
from tortoise.api import TextToSpeech, MODELS_DIR
from tortoise.utils.audio import load_voices, get_voices
-import gradio as gr
-from src.tortoise.TortoiseOutputRow import TortoiseOutputRow, TortoiseOutputUpdate
from src.tortoise.save_json import save_json
from scipy.io.wavfile import write as write_wav
from src.tortoise.TortoiseParameters import TortoiseParameters
@@ -25,6 +24,18 @@
TORTOISE_LOCAL_MODELS_DIR = get_path_from_root("data", "models", "tortoise")
+class TortoiseOutputUpdate:
+ def __init__(
+ self,
+ audio,
+ bundle_name,
+ params,
+ ):
+ self.audio = audio
+ self.bundle_name = bundle_name
+ self.params = params
+
+
def get_model_list():
try:
return ["Default"] + [
@@ -58,7 +69,7 @@ def switch_model(
tokenizer_path=tokenizer.name if tokenizer else None,
tokenizer_basic=use_basic_cleaners,
)
- return gr.Dropdown.update()
+ return gr.Dropdown()
def get_voice_list():
@@ -68,6 +79,7 @@ def get_voice_list():
def save_wav_tortoise(audio_array, filename):
write_wav(filename, SAMPLE_RATE, audio_array)
+
def unload_tortoise_model():
global MODEL
if MODEL is not None:
@@ -75,6 +87,7 @@ def unload_tortoise_model():
torch_clear_memory()
MODEL = None
+
def get_tts(
models_dir=MODELS_DIR,
force_reload=False,
@@ -87,7 +100,11 @@ def get_tts(
):
global MODEL
if MODEL is None or force_reload:
+ print("Loading tortoise model: ", models_dir)
+ print("Clearing memory...")
unload_tortoise_model()
+ print("Memory cleared")
+ print("Loading model...")
MODEL = TextToSpeech(
models_dir=models_dir,
kv_cache=kv_cache,
@@ -97,22 +114,39 @@ def get_tts(
tokenizer_vocab_file=tokenizer_path,
tokenizer_basic=tokenizer_basic,
)
+ print("Model loaded")
return MODEL
+last_voices = None
+voice_samples = None
+conditioning_latents = None
+
+
+def get_voices_cached(voice):
+ global last_voices, voice_samples, conditioning_latents
+
+ if voice == last_voices:
+ last_voices = voice
+ return voice_samples, conditioning_latents
+
+ voices = voice.split("&") if "&" in voice else [voice]
+
+ voice_samples, conditioning_latents = load_voices(
+ voices, extra_voice_dirs=[TORTOISE_VOICE_DIR]
+ )
+ last_voices = voices
+ return voice_samples, conditioning_latents
+
+
def generate_tortoise(
params: TortoiseParameters,
text: str,
candidates: int,
):
- voice = params.voice
-
os.makedirs(OUTPUT_PATH, exist_ok=True)
- voice_sel = voice.split("&") if "&" in voice else [voice]
- voice_samples, conditioning_latents = load_voices(
- voice_sel, extra_voice_dirs=[TORTOISE_VOICE_DIR]
- )
+ voice_samples, conditioning_latents = get_voices_cached(params.voice)
tts = get_tts()
result, state = tts.tts_with_preset(
@@ -160,6 +194,7 @@ def _process_gen(candidates, audio_array, id, params: TortoiseParameters):
"_type": model,
"date": date,
"candidates": candidates,
+ "index": id if isinstance(id, int) else 0,
**params.to_metadata(),
}
@@ -169,13 +204,8 @@ def _process_gen(candidates, audio_array, id, params: TortoiseParameters):
return TortoiseOutputUpdate(
audio=(SAMPLE_RATE, audio_array),
- image=filename_png,
- save_button=gr.Button.update(value="Save to favorites", visible=True),
- seed=params.seed,
bundle_name=history_bundle_name_data,
- params=gr.JSON.update(
- value=metadata
- ), # broken because gradio returns only __type__
+ params=gr.JSON(value=metadata), # broken because gradio returns only __type__
)
@@ -194,9 +224,7 @@ def get_filenames(base_filename):
return filename, filename_png, filename_json
-def generate_tortoise_long(
- outs: list[TortoiseOutputRow], count: int, params: TortoiseParameters
-):
+def generate_tortoise_long(count: int, params: TortoiseParameters):
print("Generating tortoise with params:")
print(params)
prompt_raw = params.text
@@ -211,14 +239,8 @@ def generate_tortoise_long(
text=prompt,
candidates=count,
)
- for i, data in enumerate(datas):
- yield {
- outs[i].audio: data.audio,
- outs[i].image: data.image,
- outs[i].save_button: data.save_button,
- outs[i].seed: data.seed,
- outs[i].bundle_name: data.bundle_name,
- }
+ for data in datas:
+ yield [data.audio, data.bundle_name, data.params]
for i in range(count):
audio_array = datas[i].audio[1]
@@ -226,18 +248,14 @@ def generate_tortoise_long(
# if there is only one prompt, then we don't need to concatenate
if len(prompts) == 1:
+ # return [None, None, None]
return {}
for i in range(count):
res = _process_gen(
count, np.concatenate(audio_pieces[i]), id=f"_long_{str(i)}", params=params
)
- yield {
- outs[i].audio: res.audio,
- outs[i].image: res.image,
- outs[i].save_button: res.save_button,
- outs[i].seed: res.seed,
- outs[i].bundle_name: res.bundle_name,
- }
+ yield [res.audio, res.bundle_name, res.params]
+ # return [None, None, None]
return {}
diff --git a/src/tortoise/generation_tab_tortoise.py b/src/tortoise/generation_tab_tortoise.py
index ad7fc7ca..4bf90c6b 100644
--- a/src/tortoise/generation_tab_tortoise.py
+++ b/src/tortoise/generation_tab_tortoise.py
@@ -1,9 +1,5 @@
-from typing import Any
-from src.musicgen.setup_seed_ui_musicgen import setup_seed_ui_musicgen
from src.css.css import full_css
import gradio as gr
-from src.tortoise.TortoiseOutputRow import TortoiseOutputRow
-from src.tortoise.create_tortoise_output_row_ui import create_tortoise_output_row_ui
from src.tortoise.gen_tortoise import (
generate_tortoise_long,
get_voice_list,
@@ -18,33 +14,21 @@
from src.tortoise.presets import presets
from src.tortoise.gr_reload_button import gr_open_button_simple, gr_reload_button
from src.tortoise.tortoise_model_settings_ui import tortoise_model_settings_ui
+from src.utils.randomize_seed import randomize_seed
MAX_OUTPUTS = 9
def generation_tab_tortoise():
with gr.Tab("Tortoise TTS"):
- inputs, output_rows = tortoise_core_ui()
- total_columns = len(output_rows)
-
- with gr.Row():
- for i in range(total_columns):
- target_count = total_columns - i
- generate_button(
- text=f"Generate {target_count if target_count > 1 else ''}",
- count=target_count,
- variant="primary" if target_count == 1 else "secondary",
- inputs=inputs,
- output_rows=output_rows,
- total_columns=total_columns,
- )
+ tortoise_core_ui()
def tortoise_core_ui():
with gr.Row():
with gr.Column():
model = tortoise_model_settings_ui()
- with gr.Box():
+ with gr.Column():
gr.Markdown("Voice")
with gr.Row():
voice = gr.Dropdown(
@@ -57,11 +41,11 @@ def tortoise_core_ui():
TORTOISE_VOICE_DIR_ABS, api_name="tortoise_open_voices"
)
gr_reload_button().click(
- fn=lambda: gr.Dropdown.update(choices=get_voice_list()),
+ fn=lambda: gr.Dropdown(choices=get_voice_list()),
outputs=[voice],
api_name="tortoise_refresh_voices",
)
- with gr.Box():
+ with gr.Column():
gr.Markdown("Preset")
preset = gr.Dropdown(
show_label=False,
@@ -93,8 +77,11 @@ def tortoise_core_ui():
step=0.1,
interactive=False,
)
- with gr.Box():
- seed, _, link_seed_cache = setup_seed_ui_musicgen()
+ with gr.Row():
+ seed = gr.Textbox(label="Seed", value="-1")
+ CUSTOM_randomize_seed_checkbox = gr.Checkbox(
+ label="Randomize seed", value=True
+ )
split_prompt = gr.Checkbox(label="Split prompt by lines", value=False)
@@ -109,10 +96,10 @@ def tortoise_core_ui():
preset.change(
fn=lambda x: [
- num_autoregressive_samples.update(presets[x]["num_autoregressive_samples"]),
- diffusion_iterations.update(presets[x]["diffusion_iterations"]),
- cond_free.update(
- presets[x]["cond_free"] if "cond_free" in presets[x] else True
+ gr.Slider(value=presets[x]["num_autoregressive_samples"]),
+ gr.Slider(value=presets[x]["diffusion_iterations"]),
+ gr.Checkbox(
+ value=presets[x]["cond_free"] if "cond_free" in presets[x] else True
),
],
inputs=[preset],
@@ -144,58 +131,62 @@ def tortoise_core_ui():
)
)
- with gr.Row():
- output_rows = [create_tortoise_output_row_ui(i) for i in range(MAX_OUTPUTS)]
-
- link_seed_cache(seed_cache=output_rows[0][2])
-
- return inputs, output_rows
-
+ with gr.Column():
+ audio = gr.Audio(
+ type="filepath", label="Generated audio", elem_classes="tts-audio"
+ )
+ bundle_name = gr.Textbox(
+ visible=False,
+ )
+ params = gr.JSON(
+ visible=False,
+ )
+ with gr.Row():
+ from src.history_tab.save_to_favorites import save_to_favorites
-def generate_button(text, count, variant, inputs, output_rows, total_columns):
- get_all_components = lambda count: [i for i, _, _ in output_rows[:count]]
- get_output_list = lambda count: sum(get_all_components(count), [])
- get_all_outs = lambda count: [
- TortoiseOutputRow.from_list(i) for i in get_all_components(count)
- ]
- show = lambda count: [
- gr.Column.update(visible=count > i) for i in range(total_columns)
- ]
+ gr.Button("Save to favorites").click(
+ fn=save_to_favorites,
+ inputs=[bundle_name],
+ )
- def hide_all_save_buttons(list_of_outs: list[TortoiseOutputRow]):
- return lambda: {
- outs.save_button: gr.Button.update(visible=False) for outs in list_of_outs
- }
+ def generate_button(count):
- output_cols: list[Any] = [col for _, col, _ in output_rows]
+ def gen(*args):
+ yield from generate_tortoise_long(
+ count,
+ TortoiseParameters.from_list(list(args)),
+ )
- def gen(*args):
- yield from generate_tortoise_long(
- get_all_outs(count),
- count,
- TortoiseParameters.from_list(list(args)),
+ return (
+ gr.Button(
+ value=f"Generate {count if count > 1 else ''}",
+ variant="primary" if count == 1 else "secondary",
+ )
+ .click(
+ fn=randomize_seed,
+ inputs=[seed, CUSTOM_randomize_seed_checkbox],
+ outputs=[seed],
+ )
+ .then(
+ fn=gen,
+ inputs=inputs,
+ outputs=[audio, bundle_name, params],
+ api_name=f"generate_tortoise_{count}",
+ )
)
- return (
- gr.Button(text, variant=variant)
- .click(fn=lambda: show(count), outputs=output_cols)
- .then(
- fn=hide_all_save_buttons(get_all_outs(count)),
- outputs=get_output_list(count),
- )
- .then(
- fn=gen,
- inputs=inputs,
- outputs=get_output_list(count),
- api_name=f"generate_tortoise_{count}",
- )
- )
+ with gr.Row():
+ total_columns = MAX_OUTPUTS
+ for i in range(total_columns):
+ generate_button(total_columns - i)
if __name__ == "__main__":
+ if "demo" in locals():
+ demo.close() # type: ignore
with gr.Blocks(css=full_css) as demo:
generation_tab_tortoise()
demo.launch(
- enable_queue=True,
+ server_port=7770,
)
diff --git a/src/tortoise/tortoise_model_settings_ui.py b/src/tortoise/tortoise_model_settings_ui.py
index ec98be83..75fb4979 100644
--- a/src/tortoise/tortoise_model_settings_ui.py
+++ b/src/tortoise/tortoise_model_settings_ui.py
@@ -20,7 +20,7 @@ def tortoise_model_settings_ui_inner():
TORTOISE_LOCAL_MODELS_DIR, api_name="tortoise_open_models"
)
gr_reload_button().click(
- fn=lambda: gr.Dropdown.update(choices=get_model_list()),
+ fn=lambda: gr.Dropdown(choices=get_model_list()),
outputs=[model],
api_name="tortoise_refresh_models",
)
@@ -67,7 +67,7 @@ def tortoise_model_settings_ui_inner():
def tortoise_model_settings_ui():
- with gr.Box():
+ with gr.Column():
gr.Markdown("Model")
with gr.Column():
return tortoise_model_settings_ui_inner()
diff --git a/src/utils/pip_install.py b/src/utils/pip_install.py
index 7692dc4a..545cb28c 100644
--- a/src/utils/pip_install.py
+++ b/src/utils/pip_install.py
@@ -11,9 +11,64 @@ def fn():
return fn
+def pip_uninstall_wrapper(package_name, name):
+ def fn():
+ output = []
+ for line in pip_uninstall(package_name, name):
+ output.append(line)
+ yield "
".join(output)
+
+ return fn
+
+
def pip_install(requirements, name):
+ # process = subprocess.Popen(
+ # f"pip install {requirements}",
+ # shell=True,
+ # stdout=subprocess.PIPE,
+ # stderr=subprocess.STDOUT,
+ # universal_newlines=True,
+ # )
+
+ # # Stream the output to the console
+ # for line in process.stdout: # type: ignore
+ # print(line, end="")
+ # yield line
+
+ # # Wait for the process to finish
+ # process.wait()
+
+ # # Check if the process was successful
+ # if process.returncode == 0:
+ # print(f"Successfully installed {name}")
+ # yield f"Successfully installed {name}, please restart the webui"
+ # else:
+ # print(f"Failed to install {name}")
+ # yield f"Failed to install {name}"
+ try:
+ print(f"Installing {name} dependencies...")
+ yield from pip_shell(f"pip install {requirements}")
+ print(f"Successfully installed {name} dependencies")
+ yield f"Successfully installed {name} dependencies"
+ except Exception:
+ print(f"Failed to install {name} dependencies")
+ yield f"Failed to install {name} dependencies"
+
+
+def pip_uninstall(package_name, name):
+ try:
+ print(f"Uninstalling {name} ({package_name})...")
+ yield from pip_shell(f"pip uninstall -y {package_name}")
+ print(f"Successfully uninstalled {name} ({package_name})")
+ yield f"Successfully uninstalled {name} ({package_name})"
+ except Exception:
+ print(f"Failed to uninstall {name} ({package_name})")
+ yield f"Failed to uninstall {name} ({package_name})"
+
+
+def pip_shell(command):
process = subprocess.Popen(
- f"pip install {requirements}",
+ command,
shell=True,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
@@ -28,10 +83,5 @@ def pip_install(requirements, name):
# Wait for the process to finish
process.wait()
- # Check if the process was successful
- if process.returncode == 0:
- print(f"Successfully installed {name}")
- yield f"Successfully installed {name}, please restart the webui"
- else:
- print(f"Failed to install {name}")
- yield f"Failed to install {name}"
+ if process.returncode != 0:
+ raise Exception(f"Failed to run {command}")
diff --git a/src/utils/randomize_seed.py b/src/utils/randomize_seed.py
new file mode 100644
index 00000000..9b4ac55e
--- /dev/null
+++ b/src/utils/randomize_seed.py
@@ -0,0 +1,8 @@
+import numpy as np
+
+
+def randomize_seed(seed, randomize_seed):
+ if randomize_seed:
+ return np.random.randint(0, 2**32 - 1, dtype=np.uint32)
+ else:
+ return int(seed)
diff --git a/src/utils/torch_clear_memory.py b/src/utils/torch_clear_memory.py
index 344680a1..0d83d8c8 100644
--- a/src/utils/torch_clear_memory.py
+++ b/src/utils/torch_clear_memory.py
@@ -1,6 +1,4 @@
import torch
-
-
import gc
@@ -10,4 +8,4 @@ def torch_clear_memory():
torch.cuda.empty_cache()
except Exception:
pass
- gc.collect()
\ No newline at end of file
+ gc.collect()
diff --git a/src/vocos/vocos_tab_wav.py b/src/vocos/vocos_tab_wav.py
index a8b33c05..48af24cb 100644
--- a/src/vocos/vocos_tab_wav.py
+++ b/src/vocos/vocos_tab_wav.py
@@ -36,7 +36,7 @@ def vocos_tab_wav():
file_input = gr.Audio(
label="Input Audio",
type="filepath",
- source="upload",
+ sources="upload",
interactive=True,
)