search

Rudrabha / Wav2Lip

This repository contains the codes of "A Lip Sync Expert Is All You Need for Speech to Lip Generation In the Wild", published at ACM Multimedia 2020. For HD commercial model, please try out Sync Labs
https://synclabs.so
9.8k stars 2.13k forks source link

图片人像边框 #184

Closed wowopc closed 3 years ago

wowopc commented 3 years ago

FO72I ZITE E H$T6$JL(~9 Can you handle it?

wowopc commented 3 years ago

Picture human border

prajwalkr commented 3 years ago

Is this during test-time using the pre-trained model?

wowopc commented 3 years ago

on wav2lip_gan.pth training

prajwalkr commented 3 years ago

Please explain further.

wowopc commented 3 years ago

Is using a pre-trained model

prajwalkr commented 3 years ago

I do not understand what you are trying to do.

wowopc commented 3 years ago

I run cd C:\Wav2Lip &&python inference.py --checkpoint checkpoints/wav2lip_gan.pth --face video/1.mp4 --audio video/1.m4a --pads 0 00 0 0 --resize_factor 2 --nosmooth There is a border in the image How to solve the big brother

Rudrabha commented 3 years ago

You can give more padding to see if the border becomes smoother. There is no specific way to remove border artifacts. We use a rectangular face crop for feeding the network and then paste back the generated face into the same coordinates. This can ideally be solved by a face mask used to crop exactly the facial region. At the moment, we don't have a model supporting this.

wowopc commented 3 years ago

Provide more padding? Is it to run the command or something? pads 0 00 0 0 ？

wowopc commented 3 years ago

Thank you pads The latter is better