Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Frame generation 插帧 #815

Closed
kikilloy opened this issue Jan 19, 2024 · 7 comments
Closed

Frame generation 插帧 #815

kikilloy opened this issue Jan 19, 2024 · 7 comments
Labels
enhancement New feature or request

Comments

@kikilloy
Copy link

Expected behavior 预期的功能

AMD3.0的插针技术是真不错,使我1066旋转,但也要游戏支持AMD3.0(或者替换DLSS3.0),Lossless.Scaling也弄了一个插针功能(LSFG),在某些游戏用了感觉没什么延迟(驱动面板开低延迟模式,限60帧率),Magpie 是否有这类功能的开发计划?我挺期待在Magpie 能用上插针。毕竟在Lossless.Scaling用FSR1.0提升很差,不如Magpie,我个人更喜欢用Magpie

Alternative behavior (optional) 近似的功能(可选)

No response

@kikilloy kikilloy added the enhancement New feature or request label Jan 19, 2024
@hooke007
Copy link
Collaborator

hooke007 commented Jan 19, 2024

#802 (comment)

这种后处理补帧没有神秘,都是滞后帧为代价补中间帧。

@kikilloy
Copy link
Author

滞后帧我个人可以接受,反正插针一般也就在非竞技类游戏使用,延迟高可以通过开低延迟模式弥补一下,用了Lossless.Scaling的插针的确可以在一定程度上提高流畅性,拖影个人感觉不明显,但Lossless.Scaling对GPU性能花费要求过高,插针效果并不好,除非进一步降低游戏分辨率、画质,以降低游戏在GPU上的性能花费,将剩下的性能匀给插针,所以希望Magpie能做出这个,毕竟Magpie的性能花费比Lossless.Scaling好多了

@hooke007
Copy link
Collaborator

忽视劣质的画面瑕疵,抛开作者的计划不谈,当下也没有现成开放的的超高效的后处理补帧算法。当然你如果能直接PR一个现成的方案,我想作者应该不会拒绝(

@Blinue
Copy link
Owner

Blinue commented Jan 19, 2024

我不想实现补帧有两个原因:

  1. 后处理补帧很难有好效果,而且还会增加延迟。游戏引擎因为可以使用运动矢量和深度缓冲所以补帧效果更好,但归根结底,补帧所提高的帧率是假的,不要上了老黄的当。
  2. Magpie 的目的是提高游戏画质而不是性能,和 LS 不一样。

@Blinue Blinue closed this as not planned Won't fix, can't repro, duplicate, stale Jan 19, 2024
@Blinue Blinue pinned this issue Jan 21, 2024
@Blinue Blinue changed the title 插针是否可以实现? Frame generation 插帧 Jan 21, 2024
@REB-ro
Copy link

REB-ro commented Mar 17, 2024

我不想实现补帧有两个原因:

  1. 后处理补帧很难有好效果,而且还会增加延迟。游戏引擎因为可以使用运动矢量和深度缓冲所以补帧效果更好,但归根结底,补帧所提高的帧率是假的,不要上了老黄的当。
  2. Magpie 的目的是提高游戏画质而不是性能,和 LS 不一样。

Good afternoon, Blinue.

Lossless Scaling's frame generation is different from FSR3 and DLSS3, as you said it doesn't use vectors, etc., but it manages to deliver screen fluidity. I believe the dev is using a similar technique to "Black Frame Insertion" (BFI), but instead of a black/empty frame, it repeats an entire frame without interpolating with the next frame. That's why Lossless Scaling's frame generation gives a different feeling than FSR3 and DLSS3, but the fluidity remains. If Magpie can do something similar, even if only in the testing field, it would be a good thing, because I've seen people say that Magpie is already capable of using BFI.

Lossless Scaling's frame generation is better in new versions of the software, but I trust Magpie more and I think you guys can achieve the same thing, better and with more trustfully.

Thanks, Bro.

@Blinue
Copy link
Owner

Blinue commented Mar 18, 2024

From what I understand, Lossless Scaling employs machine learning algorithms to anticipate intermediate frames. This likely results in a one-frame latency as it interpolates predicted frames between existing ones, thereby doubling the framerate.

Your attention and confidence in Magpie are much appreciated, yet there are no immediate plans to implement frame generation. This decision stems not only from the fact that it falls outside Magpie’s core objectives but also due to a backlog of highly requested features that take precedence. Nonetheless, should there be community contributions, I would gladly consider them for review.

@Linx-ESP
Copy link

Linx-ESP commented Jul 2, 2024

Frame doubling would help some 30fps games/emulators with framepacing in freesync displays (since freesymc doesnt trigger under 45Hz in some displays)

This would help as a 50 to 25 fps would be seen from.the monitor as 100 to 50 and be kept with freesync

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

5 participants