SongGen
Fully open-source single-stage auto-regressive transformer for text-to-song generation.
About
SongGen is a fully open-source, single-stage auto-regressive transformer for text-to-song generation, introduced in a February 2025 paper. It turns lyrics plus a text description into songs with harmonized vocals and accompaniment, with fine-grained control over instruments, genre, mood, and timbre, and supports zero-shot voice cloning from a three-second reference clip.
It offers two output modes: a mixed mode blending vocals and accompaniment into one track, and a dual-track mode synthesizing them separately for professional post-production. The authors plan to release weights, training code, data, and the preprocessing pipeline; the voice-cloning capability carries the usual misuse risks, so responsible deployment matters.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: Paper-stage model; community adoption still early.; Provider institution not confirmed from sources; zero-shot voice cloning carries misuse risk.. License: Described as fully open-source in the paper; confirm exact license on release before commercial use..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 4.0/10
- Maintenance · recency, releases, issue load
- 4.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · Paper-stage model; community adoption still early.
- · Provider institution not confirmed from sources; zero-shot voice cloning carries misuse risk.