ClearerVoice-Studio
github.com
What it is
ClearerVoice-Studio is an open-source toolkit for AI speech processing, aimed at researchers, developers, and end-users. Its tasks include speech enhancement, speech separation, speech super-resolution, and audio-visual target-speaker extraction. The project provides pretrained models for these tasks and scripts for training and fine-tuning. It also includes SpeechScore for intrusive and non-intrusive speech-quality assessment, with metrics including PESQ, DNSMOS, SNR, and SI-SDR. The toolkit processes audio and video in batches and provides a Python interface for file paths, directories, `.scp` lists, and NumPy input-to-output use. Listed audio formats include WAV, AAC, AC3, AIFF, FLAC, M4A, MP3, OGG, OPUS, WMA, and WEBM; FFmpeg is required for codecs beyond WAV. Installation is available through `pip install clearvoice` or from the GitHub repository. Pretrained models are fetched from Hugging Face during inference, with ModelScope downloads available manually. It is released under Apache License, Version 2.0.
Spec sheet
- Purpose
- ClearerVoice-Studio is an open-source AI speech-processing toolkit with pretrained models for speech enhancement, separation, and target speaker extraction.Read 9 Oct 2026 · github.com
- Intended users
- The repository describes the toolkit as designed for researchers, developers, and end-users.Read 9 Oct 2026 · github.com
- Platforms
- The installation instructions describe FFmpeg setup for Ubuntu/Debian, macOS, and Windows.Read 4 Oct 2026 · github.com
- Support
- The maintainers invite users to raise repository issues, email the listed contacts, or join their DingTalk group.Read 9 Oct 2026 · github.com
- License
- The GitHub repository identifies its license as Apache-2.0.Read 9 Oct 2026 · github.com
- Additional task
- ClearVoice also provides speech super-resolution and audio-visual target speaker extraction.Read 9 Oct 2026 · github.com
- Audio formats
- The repository says ClearVoice supports formats including WAV, AAC, AC3, AIFF, FLAC, M4A, MP3, OGG, OPUS, WMA, and WEBM, and requires FFmpeg for audio codecs.Read 9 Oct 2026 · github.com
- Demos
- The repository links to ClearVoice demos on Hugging Face and ModelScope, plus a SpeechScore demo on Hugging Face.Read 4 Oct 2026 · github.com
- Installation
- ClearVoice can be installed from PyPI with `pip install clearvoice` or installed from the GitHub repository.Read 9 Oct 2026 · github.com
- Integration
- The README shows a Python interface that accepts audio or video file paths, directories, or `.scp` lists, and includes a NumPy input-to-output interface.Read 9 Oct 2026 · github.com
- Model downloads
- Pretrained models are fetched from Hugging Face during inference, with manual downloads available from ModelScope if automatic downloads fail.Read 9 Oct 2026 · github.com
- Model hosting
- Pretrained models are fetched from Hugging Face during inference, with ModelScope downloads offered as a manual alternative.Read 4 Oct 2026 · github.com
- Notable limit
- The README says Hugging Face demo GPU usage is limited and ModelScope has a larger GPU usage quota.Read 4 Oct 2026 · github.com
- Pretrained models
- The listed pretrained models cover speech enhancement, speech separation, speech super-resolution, and audio-visual target speaker extraction.Read 9 Oct 2026 · github.com
- Quality assessment
- The included SpeechScore toolkit evaluates speech quality using metrics including SNR, PESQ, STOI, DNSMOS, and SI-SDR.Read 9 Oct 2026 · github.com
- Speech tasks
- ClearVoice supports speech enhancement, speech separation, speech super-resolution, and audio-visual target-speaker extraction.Read 4 Oct 2026 · github.com
- Training
- The repository provides training and fine-tuning scripts for speech enhancement, separation, super-resolution, and target speaker extraction.Read 9 Oct 2026 · github.com
Strong points and limits
Strong points
- Four speech-processing task types
- Includes pretrained models and training scripts
- SpeechScore offers multiple assessment metrics
- Audio and video batch processing
- Apache License, Version 2.0
Limits
- FFmpeg required for audio codecs beyond WAV
- Pretrained models are fetched from Hugging Face
- Hugging Face demo GPU usage is limited
Who it suits
ClearerVoice-Studio suits researchers and developers who need speech-processing models, training scripts, or quality assessment. End-users can also use its processing toolkit and linked demos.
Verdict
ClearerVoice-Studio combines speech processing, model training, and speech-quality assessment in one open-source toolkit. Account for its model-download setup and the FFmpeg requirement for non-WAV codecs.
Where Specifiction places it
- Runtime Version Managers#14 of 307.6
- AI Speech Enhancement Software#20 of 357.3
