DDSP-SVC
DDSP-SVC is a free, open-source project for converting singing voices in real time on a personal computer.
What DDSP-SVC is
DDSP-SVC is a free, open-source project for converting singing voices in real time on a personal computer. It runs locally and supports Windows; its README describes installing Python dependencies and launching a conversion GUI. The real-time interface uses sliding windows, cross-fading, SOLA-based splicing, and contextual semantic reference. For offline work, a command-line interface converts an input WAV to an output WAV, with pitch-shift and speaker options. Users can prepare audio datasets and train models with a rectified-flow script, including multi-speaker models configured with a speaker count and speaker-numbered folders. Setup instructions support ContentVec or HubertSoft encoders, an NSF-HiFiGAN vocoder, and the RMVPE pitch extractor. The project recommends PyTorch and documents TensorBoard for viewing training progress and generated test audio. Its default configuration is described for 44.1 kHz training with an RTX 4060. Dataset guidance suggests about 1,000 training clips and about 10 validation clips, each at least two seconds long. The author asks users to use legally obtained, authorized data and avoid illegal uses of models or generated audio.
Good
- Free and open source
- Real-time and offline WAV conversion
- Supports pitch-shift and speaker options
- Training supports multi-speaker models
- Runs locally on Windows
Watch out
- Requires local Python dependencies and model assets
- Dataset preparation and model training require setup
- Author restricts training to authorized data
Who it suits
DDSP-SVC suits users who want local singing voice conversion and are prepared to set up Python dependencies and model assets. It also supports users who want to train single- or multi-speaker models from audio datasets.
Verdict
DDSP-SVC combines local real-time conversion with offline WAV conversion and model training. It is a fit for users comfortable with its Python setup and dataset preparation requirements.
Plans and prices
From DDSP-SVC's own pricing page, checked 9 Oct 2026.
| Plan | Price | What it gives you |
|---|---|---|
| Free | Free | Personal computer software · Requires local Python dependencies and pretrained model assets |
The facts
Each one read from DDSP-SVC's own pages, with the date it was checked.
What it does
- Data and legal use
- The author asks users to train only with legally obtained authorized data and not use models or generated audio for illegal purposes.checked 9 Oct 2026
- Dataset guidance
- The project recommends about 1,000 training clips and about 10 validation clips, with each audio clip at least two seconds long.checked 9 Oct 2026
- Hardware and setup
- The README says its default configuration suits 44.1 kHz training with an RTX 4060 and reports a working Python 3.11, Windows, CUDA 13.0, PyTorch 2.9.1, and torchaudio 2.9.1 setup.checked 9 Oct 2026
- Integrations
- The README recommends installing PyTorch and documents TensorBoard for viewing training status and generated test audio.checked 9 Oct 2026
- Intended users
- The project describes its goal as developing free AI voice-changer software that can be used on personal computers.checked 9 Oct 2026
- Local processing
- The README describes installing dependencies and running the conversion GUI locally with Python.checked 9 Oct 2026
- Model components
- The setup instructions support ContentVec or HubertSoft encoders, an NSF-HiFiGAN vocoder, and the RMVPE pitch extractor.checked 9 Oct 2026
- Model training
- The project documents preprocessing audio datasets and training a model with a rectified-flow script.checked 9 Oct 2026
- Multi-speaker support
- Training supports multi-speaker models configured with a speaker count and speaker-numbered audio folders.checked 9 Oct 2026
- Offline conversion
- The command-line interface converts an input WAV file to an output WAV file and accepts pitch-shift and speaker options.checked 9 Oct 2026
- Real-time conversion
- Its real-time interface uses sliding windows, cross-fading, SOLA-based splicing, and contextual semantic reference.checked 9 Oct 2026
- What it does
- DDSP-SVC is an open-source project for real-time singing voice conversion.checked 9 Oct 2026
Alternatives to DDSP-SVC
- 1
BeMusic AI Cover
Free planAndroidiOSWebLinux8.2From $12 a month - 2
Revocalize AI
Free planAndroidiOSWebAPI8.1From $9 a month - 3
Media.io AI Song Cover Generator
Free planAndroidiOSWebAPI8.0From $59.90 a year - 4
Kits AI
Free planWebAPIWindows7.7From $10 a month - 5
Applio
Free planOpen sourceWebAPILinuxmacOS7.6Free plan - 6
Audjust AI Song Cover Generator
Free planWebAPI7.3From $10 a month