High-Quality Automatic Voice Over with Accurate Alignment: Supervision through Self-Supervised Discrete Speech Units
This repository contains the source code and speech samples for the paper accepted to INTERSPEECH 2023 by Junchen Lu, Berrak Sisman, Mingyang Zhang, Haizhou Li.
The goal of Automatic Voice Over (AVO) is to generate speech in sync with a silent video given its text script. Recent AVO frameworks built upon text-to-speech synthesis (TTS) have shown impressive results. However, the current AVO learning objective of acoustic feature reconstruction brings in indirect supervision for inter-modal alignment learning, thus limiting the synchronization performance and synthetic speech quality. To this end, we propose a novel AVO method leveraging the learning objective of self-supervised discrete speech unit prediction, which not only provides more direct supervision for the alignment learning, but also alleviates the mismatch between the text-video context and acoustic features. Experimental results show that our proposed method achieves remarkable lip-speech synchronization and high speech quality by outperforming baselines in both objective and subjective evaluations.
Voice over video samples are available here.
- Python >= 3.6
- PyTorch v1.7
- Install dependencies
git clone https://github.com/ranacm/DSU-AVO.git cd DSU-AVO pip install -r requirements.txt
- The full list of Chem dataset can be found at Lip2Wav dataset. Download and preprocess the dataset follow Neural Dubber paper. For preprocessed data, please send an inquiry to the author at [email protected].
- To quantize the speech audio data with HuBERT, change the first line of
hubert_tokenizer/chem_manifest.txt
to audio dataset path, put the pretrained model weights underhubert_tokenizer/pretrained_models
and use the commands provided below. More details and pretrained quantizer models can be found at GSLM code.cd hubert_tokenizer python quantize_with_kmeans.py \ --feature_type hubert \ --kmeans_model_path ../hubert_tokenizer/pretrained_models/km.bin \ --acoustic_model_path ../hubert_tokenizer/pretrained_models/hubert_base_ls960.pt \ --layer 6 \ --manifest_path ../hubert_tokenizer/chem_manifest.txt \ --out_quantized_file_path Chem_hubert100.txt \ --extension ".wav"
- For visual data preparation, follow the instructions described in AV-HuBERT preparation.
- To train the unit vocoder with Chem dataset follow the instructions described in Speech Resynthesis. Unit vocoder config is provided at
unit_vocoder/Chem_hubert100.json
. The pretrained model can be downloaded here.
This repository is currently under construction. The source code for experiments described in our paper will be made publicly available very soon. We appreciate your patience and interest in our work.
If you find our work interesting, please consider citing our paper:
@inproceedings{lu23f_interspeech,
author={Junchen Lu and Berrak Sisman and Mingyang Zhang and Haizhou Li},
title={{High-Quality Automatic Voice Over with Accurate Alignment: Supervision through Self-Supervised Discrete Speech Units}},
year=2023,
booktitle={Proc. INTERSPEECH 2023},
pages={5536--5540},
doi={10.21437/Interspeech.2023-2179}
}