search

rlgnswk / NeRFFaceSpeech_Code

One-shot Audio-driven 3D Talking Head Synthesis via Generative Prior, CVPRW 2024
https://rlgnswk.github.io/NeRFFaceSpeech_ProjectPage/
MIT License
53 stars 2 forks source link

After post processing the results are slightly improved #5

Closed nitinmukesh closed 1 month ago

nitinmukesh commented 1 month ago

https://github.com/user-attachments/assets/a480f936-a376-4cf3-92b2-5c97e45c9fab

Left is the output from tool, and right is after gfpgan.

The pipeline can be included in the tool easily and code is available in the following repos. https://github.com/X-LANCE/AniTalker https://github.com/tanshuai0219/EDTalk

rlgnswk commented 1 month ago

Thank you for deriving additional results!

I recognize that existing research suggests that post-processing can enhance the final quality. However, since this is not the method proposed in this paper, I believe it should not be included.

Nonetheless, I really appreciate that you demonstrated this through experiments.

If it’s alright, I would like to mention this issue in the Readme to inform others about this method.

nitinmukesh commented 1 month ago

Yes, please you can mention in Readme.

I will do more test with inference methods tomorrow.

rlgnswk commented 1 month ago

I've updated it! Thank you for making the code more complete.