Sample outputs

[NikitaKononov]

Sample outputs

[cmmclee]

Thanks for your quick response

sorry, i can't share but you can see the issue: #10 (comment) or #10 (comment)

[NikitaKononov]

sorry, i can't share but you can see the issue: #10 (comment)

Do the results in this issue have the same visual quality as your model? Or you have a way better result?
Why you can't btw?

sorry because I left the organization, this is my commitment. This is nearly my result

[NikitaKononov]

sorry because I left the organization, this is my commitment. This is nearly my result

Understood, thanks
Hm.. Result is not very impressive to be honest, much artefacts
Do you know any modern models for lip-sync or talking head generation?
I've tried FACIAL, but it's blinking quality and overall visual quality isn't very good

you can contact the author wav2lip to see the result 288x192, this result is very good. But this implementation has a lot of limitation. because it just gets 5 frames and I saw it doesn't has any connection from the previous step to the next step. So I think you need to implement better model through LSTM or RNN or something can make connection

[NikitaKononov]

you can contact the author wav2lip to see the result 288x192, this result is very good. But this implementation has a lot of limitation. because it just gets 5 frames and I saw it doesn't has any connection from the previous step to the next step. So I think you need to implement better model through LSTM or RNN or something can make connection

https://iiitaphyd-my.sharepoint.com/personal/radrabha_m_research_iiit_ac_in/_layouts/15/onedrive.aspx?id=%2Fpersonal%2Fradrabha_m_research_iiit_ac_in%2FDocuments%2FWav2Lip_results%2FSravan%2FSravan_Telugu_720p%2Emp4&parent=%2Fpersonal%2Fradrabha_m_research_iiit_ac_in%2FDocuments%2FWav2Lip_results%2FSravan&ga=1

Their 288x192 result
Can't call it very good, especially for the price of 7500$

[cmmclee]

Hello @NikitaKononov This video address cannot be accessed. Can you upload this video file? Thanks

[NikitaKononov]

Hello @NikitaKononov This video address cannot be accessed. Can you upload this video file? Thanks

Sravan_Telugu_720p.mp4

My method can get better result but I can take a lot of resources gpu

[NikitaKononov]

My method can get better result but I can take a lot of resources gpu

Alot of resources while training? or while inference
will you shade your method?

I mean this repo