SoulX-Singer
github.com
- Free plan
- Linux
- Web
- Self-hosted
What SoulX-Singer is
SoulX-Singer is an open-source model for generating singing voices for singers it has not been tuned on. Users can control pitch, rhythm, and expression through melody-conditioned F0 contours or score-conditioned MIDI notes. Its voice-conversion system changes raw singing audio to a target singer’s voice while retaining melody, rhythm, and lyrics, without requiring lyric or MIDI transcriptions. The system supports Mandarin Chinese, English, and Cantonese, along with singer-timbre cloning, cross-lingual synthesis, and lyric editing while preserving natural prosody. Preprocessing tools handle vocal separation, dereverberation, F0 extraction, voice activity detection, and lyric and note transcription. A MIDI editor can adjust lyrics, phoneme alignment, pitches, and durations before synthesis. Soul-AILab provides singing-generation and vocal-conversion demos, plus a separate MIDI editor, through Hugging Face Spaces. The repository supports local inference using Conda, Python 3.10, pip-installed dependencies, and web UI scripts. The project reports more than 42,000 hours of aligned vocal, lyric, and note data. Code and weights use the Apache 2.0 license. Maintainers caution that automatic preprocessing may misalign singing with lyrics and notes, so manual correction may be needed.
Who it suits
SoulX-Singer may suit researchers and developers working on singing synthesis or voice conversion. Its web demos offer online access, while the repository supports local inference.
Good to know, and what to watch
- Generates voices for singers without per-speaker fine-tuning
- Controls pitch, rhythm, and expression with F0 or MIDI
- Voice conversion preserves melody, rhythm, and lyrics
- Supports Mandarin Chinese, English, and Cantonese
- Apache 2.0 code and model weights
- Automatic preprocessing may misalign singing with lyrics and notes
- Local inference requires Conda, Python 3.10, and dependencies
- Use must respect intellectual property, privacy, and consent
Plans
Every plan and priceSoulX-Singer
Open source
Getting started
- Open the project repository or a Hugging Face Space
- Choose the singing-generation, vocal-conversion or MIDI-editor demo
- For local use, clone the repository and create the Conda environment
- Install the pip dependencies with Python 3.10
- Download synthesis, conversion and preprocessing models through Hugging Face Hub commands
- Run the local web UI scripts
Limits to check
The maintainers warn that automatic preprocessing may misalign singing audio with lyrics and notes, so manual correction can be required. Usage must respect intellectual property, privacy and consent and cannot be used for unauthorized impersonation or deceptive audio.
The verdict
SoulX-Singer offers synthesis and conversion controls, MIDI editing, web demos, and local inference. Automatic preprocessing can misalign material, so users may need to correct it manually.
Questions people ask
Is SoulX-Singer free?
Yes. Code and model weights are free to use under the Apache 2.0 license.
Which languages are supported?
The system supports Mandarin Chinese, English and Cantonese.
Can it convert an existing vocal recording?
Yes. SoulX-Singer-SVC converts raw singing audio into a target singer’s voice while preserving melody, rhythm and lyrics without lyric or MIDI transcriptions.
Does it support MIDI editing?
Yes. MIDI metadata can be edited for lyrics, phoneme alignment, pitches and durations, then imported back for synthesis.
Where can it run?
SoulX-Singer supports Linux, self-hosted deployment and web access through Hugging Face Spaces. Local inference uses Conda, Python 3.10, pip dependencies and web UI scripts.
Are there usage restrictions?
The maker asks users to respect intellectual property, privacy and consent and prohibits unauthorized impersonation or deceptive audio.
What the maker publishes
- Control modes
- It supports melody-conditioned F0 contour control and score-conditioned MIDI note control for pitch, rhythm, and expression. source
- Core function
- SoulX-Singer is a high-fidelity zero-shot singing voice synthesis model for generating realistic voices for unseen singers. source
- Dataset scale
- The stated training dataset contains more than 42,000 hours of aligned vocals, lyrics, and notes. source
- Deployment
- The repository can be cloned, installed with Conda and pip, and run locally through Python web UI scripts. source
- Editing and cloning
- Features include singer-timbre cloning, cross-lingual synthesis, and lyric editing while preserving natural prosody. source
- Languages
- The system supports Mandarin Chinese, English, and Cantonese. source
- License
- The code and model weights are released under the Apache 2.0 license for researchers and developers to use. source
- License and safety
- The project uses Apache 2.0 and asks users to respect intellectual property, privacy, and consent and avoid unauthorized impersonation or deceptive audio. source
- Local deployment
- The repository supports local inference through Conda with Python 3.10, pip-installed dependencies, and local WebUI scripts. source
- MIDI editing
- A MIDI Editor supports editing lyrics, phoneme alignment, note pitches, and durations before inference. source
- MIDI integration
- Generated metadata can be exported to MIDI, edited for lyrics, phoneme alignment, pitches, and durations, and imported back for synthesis. source
- Model distribution
- Pretrained synthesis, conversion, and preprocessing models are downloaded through Hugging Face Hub commands. source
- Notable limitation
- The maintainers warn that automatic preprocessing may misalign singing audio with lyrics and notes and recommend manual correction. source
- Online access
- Soul-AILab provides a running SoulX-Singer demo on Hugging Face Spaces and a separate running MIDI Editor Space. source
- Pitch and score control
- It supports melody-conditioned F0-contour control and score-conditioned MIDI-note control for pitch, rhythm, and expression. source
- Preprocessing
- Its preprocessing toolkit performs vocal separation and dereverberation, F0 extraction, voice activity detection, lyrics transcription, and note transcription. source
- Purpose
- SoulX-Singer is a high-fidelity zero-shot singing voice synthesis model for generating realistic voices for unseen singers. source
- Support
- The project lists three contact emails and invites technical discussion through WeChat or Soul app groups. source
- Training data
- The project reports more than 42,000 hours of aligned vocal, lyric, and note data. source
- Usage restrictions
- The maker asks users to respect intellectual property, privacy, and consent and prohibits unauthorized impersonation or deceptive audio. source
- Voice conversion
- SoulX-Singer-SVC converts raw singing audio into a target singer’s voice while preserving melody, rhythm, and lyrics without lyric or MIDI transcriptions. source
- Web access
- The maker provides a SoulX-Singer singing-generation and vocal-conversion demo on Hugging Face Spaces. source
- Zero-shot operation
- The model generates voices for unseen singers without fine-tuning or per-speaker fine-tuning. source
Checked against the maker's pages on 1 Oct 2026.