wangsuzhen / audio2head

code for paper "Audio2Head: Audio-driven One-shot Talking-head Generation with Natural Head Motion" in the conference of IJCAI 2021

Audio2Head: Audio-driven One-shot Talking-head Generation with Natural Head Motion (IJCAI 2021)

Paper | Demo

Requirements

• Python 3.6 , Pytorch >= 1.6 and ffmpeg

• Other requirements are listed in the 'requirements.txt'

Pretrained Checkpoint

Please download the pretrained checkpoint from google-drive and put it within the folder (/checkpoints).

Generate Demo Results

python inference.py --audio_path xxx.wav --img_path xxx.jpg

Note that the input images must keep the same height and width and the face should be appropriately cropped as in /demo/img.

License and Citation

@InProceedings{wang2021audio2head,
author = Suzhen Wang, Lincheng Li, Yu Ding, Changjie Fan, Xin Yu
title = {Audio2Head: Audio-driven One-shot Talking-head Generation with Natural Head Motion},
booktitle = {the 30th International Joint Conference on Artificial Intelligence (IJCAI-21)},
year = {2021},
}

Acknowledgement

This codebase is based on First Order Motion Model, thanks for their contribution.

audio2head's Issues

License

Thank you for your research.
Can i use this code and pretrained model for commercial use?
I wanna know about license.

Instructions for training a new model

Hi

How to train a new model?

How to prepare the dataset?
and how to invoke the training code? what are the optimal hyperparameters? etc.

我们创建了一个中文讨论组，有需要的加我微信douzijun1999

1705126444.mp4

issue：TypeError: write() got an unexpected keyword argument 'fps'

issue：Traceback (most recent call last):
File "inference.py", line 251, in
audio2head(parse.audio_path,parse.img_path,parse.model_path,parse.save_path)
File "inference.py", line 233, in audio2head
imageio.mimsave(video_path, predictions_gen,fps=25.0)
File "D:\miniconda3\envs\a2\lib\site-packages\imageio\v2.py", line 495, in mimwrite
return file.write(ims, is_batch=True, **kwargs)
File "D:\miniconda3\envs\a2\lib\site-packages\imageio\plugins\tifffile_v3.py", line 244, in write
self._fh.write(image, **kwargs)
TypeError: write() got an unexpected keyword argument 'fps'

RuntimeError: shape '[-1, 1, 4, 41]' is invalid for input of size 186780

RuntimeError: shape '[-1, 1, 4, 41]' is invalid for input of size 186780

已创建中文的讨论组想加入的请添加微信xaaheng

issuses TypeError: load() missing 1 required positional argument: 'Loader'

C:\Users\flyingree\Downloads\Audio2Head-main>python inference.py --audio_path temp.wav --img_path 1.jpg ffmpeg version 6.0-essentials_build-www.gyan.dev Copyright (c) 2000-2023 the FFmpeg developers built with gcc 12.2.0 (Rev10, Built by MSYS2 project) configuration: --enable-gpl --enable-version3 --enable-static --disable-w32threads --disable-autodetect --enable-fontconfig --enable-iconv --enable-gnutls --enable-libxml2 --enable-gmp --enable-bzlib --enable-lzma --enable-zlib --enable-libsrt --enable-libssh --enable-libzmq --enable-avisynth --enable-sdl2 --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxvid --enable-libaom --enable-libopenjpeg --enable-libvpx --enable-mediafoundation --enable-libass --enable-libfreetype --enable-libfribidi --enable-libvidstab --enable-libvmaf --enable-libzimg --enable-amf --enable-cuda-llvm --enable-cuvid --enable-ffnvcodec --enable-nvdec --enable-nvenc --enable-d3d11va --enable-dxva2 --enable-libvpl --enable-libgme --enable-libopenmpt --enable-libopencore-amrwb --enable-libmp3lame --enable-libtheora --enable-libvo-amrwbenc --enable-libgsm --enable-libopencore-amrnb --enable-libopus --enable-libspeex --enable-libvorbis --enable-librubberband libavutil 58. 2.100 / 58. 2.100 libavcodec 60. 3.100 / 60. 3.100 libavformat 60. 3.100 / 60. 3.100 libavdevice 60. 1.100 / 60. 1.100 libavfilter 9. 3.100 / 9. 3.100 libswscale 7. 1.100 / 7. 1.100 libswresample 4. 10.100 / 4. 10.100 libpostproc 57. 1.100 / 57. 1.100 Guessed Channel Layout for Input Stream #0.0 : mono Input #0, wav, from 'temp.wav': Duration: 00:00:10.00, bitrate: 352 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 22050 Hz, 1 channels, s16, 352 kb/s Stream mapping: Stream #0:0 -> #0:0 (pcm_s16le (native) -> pcm_s16le (native)) Press [q] to stop, [?] for help Output #0, wav, to './results/temp.wav': Metadata: ISFT : Lavf60.3.100 Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, mono, s16, 256 kb/s Metadata: encoder : Lavc60.3.100 pcm_s16le size= 313kB time=00:00:10.00 bitrate= 256.1kbits/s speed=1.5e+03x video:0kB audio:313kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.024367% torch.Size([1, 250, 6]) Traceback (most recent call last): File "C:\Users\flyingree\Downloads\Audio2Head-main\inference.py", line 251, in <module> audio2head(parse.audio_path,parse.img_path,parse.model_path,parse.save_path) File "C:\Users\flyingree\Downloads\Audio2Head-main\inference.py", line 141, in audio2head config = yaml.load(f) TypeError: load() missing 1 required positional argument: 'Loader'

So what is the matter?

How to train?

Dear suzhen,

Thank you for publishing the inference code. It is an amazing work. But how can I train a new model with higher resolution. I would like to ask whether you will publish the train code.

Looking forward to your kind reply.

Regards,
Bolin

Can generate video without head motions?

Is pretrained model trained on voxceleb1?

I've read your paper and am excited to test it myself.

However, I have a question below just for double-checking purpose :
Is the released pretrained model trained on voxceleb1?

thank you

FileNotFoundError: [Errno 2] No such file or directory: './results/temp.wav'

mac 环境，直接运行报错

python3 inference.py --audio_path input.wav --img_path input.png
/bin/sh: ffmpeg: command not found
Traceback (most recent call last):
File "/Users/xxx/Documents/GitHub_Source/Audio2Head/inference.py", line 251, in
audio2head(parse.audio_path,parse.img_path,parse.model_path,parse.save_path)
File "/Users/xxx/Documents/GitHub_Source/Audio2Head/inference.py", line 125, in audio2head
audio_feature = get_audio_feature_from_audio(temp_audio)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/xxx/Documents/GitHub_Source/Audio2Head/inference.py", line 99, in get_audio_feature_from_audio
sample_rate, audio = wavfile.read(audio_path)
^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.11/site-packages/scipy/io/wavfile.py", line 647, in read
fid = open(filename, 'rb')
^^^^^^^^^^^^^^^^^^^^
FileNotFoundError: [Errno 2] No such file or directory: './results/temp.wav'

Is that possible to generate 512X512 video?

It is real cool work! However I see it generates 256X256 video.

Is that possible to generate 512X512 video?

Thanks.