Does coqui-tts support stateful inference?

[asr-aditya]

I have implemented the xtts streaming inference. However, the generated output seems to speak each word separately when I stream input word by word without the context of the previous words. How to pass the previous state to the model for inference? So that the speech sounds natural. Does this functionality already exist? If not how can we achieve this?

import time
import torch
import torchaudio
from TTS.tts.configs.xtts_config import XttsConfig
from TTS.tts.models.xtts import Xtts

print("Loading model...")
config = XttsConfig()
config.load_json("/path/to/xtts/config.json")
model = Xtts.init_from_config(config)
model.load_checkpoint(config, checkpoint_dir="/path/to/xtts/", use_deepspeed=True)
model.cuda()

print("Computing speaker latents...")
gpt_cond_latent, speaker_embedding = model.get_conditioning_latents(audio_path=["reference.wav"])

print("Inference...")
t0 = time.time()
chunks = model.inference_stream(
    "It took me quite a long time to develop a voice and now that I have it I am not going to be silent.",
    "en",
    gpt_cond_latent,
    speaker_embedding
)

wav_chuncks = []
for i, chunk in enumerate(chunks):
    if i == 0:
        print(f"Time to first chunck: {time.time() - t0}")
    print(f"Received chunk {i} of audio length {chunk.shape[-1]}")
    wav_chuncks.append(chunk)
wav = torch.cat(wav_chuncks, dim=0)
torchaudio.save("xtts_streaming.wav", wav.squeeze().unsqueeze(0).cpu(), 24000)