Distorted Output about talking-face_pc-avs HOT 6 OPEN

Please refer to the "Self-Prepared Data Processing" part in the README to align your faces.

from talking-face_pc-avs.

Hey @Hangz-nju-cuhk
I firstly ran scripts/prepare_testing_files.py on my videos and then ran scripts/align_68.py. After this i changed the folders paths to these [name_cropped] folders in the .csv file.
Then I ran bash experiments/demo_vox.sh still I get the distorted outputs

I ran scripts/prepare_testing_files.py with the following flags --src_pose_path video_file.mp4 --src_audio_path video_file.mp4 --src_input_path test_image.jpg

Let me know if there are certain flags i have wronglt set

from talking-face_pc-avs.

Please show your current results. If the images are accurately aligned, the same region as shown in the demos should be saved. Otherwise, the alignment is not successful.

from talking-face_pc-avs.

@Hangz-nju-cuhk
I didn't align the image which was causing the issue. I got the following output. Just wondering is it possible to copy the lip/mouth movements completely from the video?

• Will it work if I provide mute audio?
• Is there some way to predict on just video?
• Any thoughts on how to model the eyes/eyebrows?

from talking-face_pc-avs.

If you would like to use videos to drive, you could try the FOMM model https://github.com/AliaksandrSiarohin/first-order-model.

from talking-face_pc-avs.

@Hangz-nju-cuhk Possible to re-train using parts of your model to achieve eye/eyebrows and lips movement?

from talking-face_pc-avs.