xinxin-song review
A focused, self-hosted pipeline for technical users who want control over cover production.
Reviewed by iTechGuides Editors · Editorial team · Updated Oct 2026
xinxin-song is a locally deployed AI song-cover pipeline for solo creators and small teams that want to convert an original song into a version sung with a specified voice. It combines Demucs-based vocal and instrumental separation with RVC-based singing-voice conversion. Users can also provide external vocal and instrumental tracks, making the workflow suitable for projects that already have separated or replacement audio. The product is command-line software rather than a hosted service, with WAV and MP3 export.
Its strongest area is control over the audio pipeline. Pitch shifting works by semitone, while configurable F0 algorithms include RMVPE, Harvest, Crepe, and PM. Extracted vocals can receive denoising and high-pass filtering before final mixing, which includes loudness and clipping protection. Optional cleanup of intermediate files keeps the local workflow more manageable. These controls make xinxin-song a fit for technically confident users who want to adjust conversion and mixing rather than work through a simplified web interface.
The trade-off is setup effort and ecosystem scope. A local Python environment, RVC models, a trained voice model, and FFmpeg are required, and voice-model training takes place separately through RVC WebUI. The published deployment options center on Windows and self-hosting, with documentation as the support channel. xinxin-song should suit users who value local processing, independent audio inputs, and configurable conversion settings. It is less suitable for buyers seeking browser-based access, built-in voice creation, broad platform coverage, or an integrated voice-model training workflow.
xinxin-song pros and cons
- Where it wins
- Combines Demucs separation and RVC singing-voice conversion
- Offers pitch, F0, denoising, filtering, and mixing controls
- Accepts separate vocal and instrumental inputs with WAV or MP3 output
- Where it doesn't
- Requires local Python, RVC models, a trained voice model, and FFmpeg
- Runs from the command line on Windows or a self-hosted setup
- Voice-model training is handled separately through RVC WebUI
xinxin-song fact sheet, pricing and score →
Advertiser disclosure: iTechGuides is reader-supported. Vendors can pay for top positions in our rankings and for a place on other products' pages, and we may earn a commission when you click some links. Bought places stand at the number bought; this page and our disclosure say so. Paying vendors may get a fuller, more detailed review from our editors, and that review can change their score. How we rank.
Last updated · How we research and update