Closed AlonDan closed 3 years ago
Is there a chance that a High Quality Model will be released soon?
Please check the first line of the README
Is there a chance that a High Quality Model will be released soon?
Please check the first line of the README
Sorry, maybe you can help because I didn't find the instructions for how to use a High Quality Model in the Readme as you suggest in the first line:
# **Wav2Lip**: *Accurately Lip-syncing Videos In The Wild*
or also later on when I tried to look for an answer to my question above, high quality model instead of the original model. Can you please tell me where it is so I can try and follow?
I tried to guess how to do it, because I really don't have a clue and there are no tutorials or good guides to explain how it's done: I tried to replace the original (older) model: 🔸wav2lip_gan.pth🔸 with the 🔹visual_quality_disc.pth🔹 but it crashes... so I'm not sure how to make it work. I just replaced it on the command, maybe I'm missing something?
How can I make it work with the 🔹visual_quality_disc.pth🔹 ? do I need to change or add something to the command?
If anyone can give an example how to use the High Quality Model by making a short step-by-step tutorial please it will be very helpful for many people who wants to experiment and test.
Thanks ahead!
@AlonDan have you figured out how to use the visual_quality_disc.pth?
@Ha0Tang Unfortunately there is no solution or support or step-by-step tutorial to explain how to make it work correct.
Many months ago I think it was mentioned that a better high-res model is on the working and will released soon.
It's been a while, and since it's very hard for newbies to understand or follow the current documents in order to train our own models, unfortunately, there are no step-by-step tutorials written or in YouTube at the moment.
I would train my own model if it wasn't so hard and complicated to understand how to do, but I tried and it's very complicated.
Is there a chance that a High Quality Model will be released soon? or higher-quality model in any way instead of the low-res current one?
Thanks ahead!