TangoFlux
515M-parameter text-to-audio via flow matching — 30 seconds of audio in under 4
About
TangoFlux is a compact text-to-audio model from a team including Stability AI researchers that trades diffusion for flow matching plus CLAP-ranked preference optimization. At 515M parameters it generates 30-second clips in about 3.7 seconds on an A40 — dramatically faster than the diffusion baselines it was compared against — while reporting the best Frechet distance, CLAP alignment, and inception scores in its comparison table (beating AudioLDM 2-large, Stable Audio Open, and Tango 2 on most metrics). The weights carry the Stability AI Community License, so use is research and non-commercial unless separately licensed, and some training data terms restrict academic use. It is a research codebase rather than a polished product, but a strong efficiency baseline.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: FD_openl3 75.1, CLAP 0.480, IS 12.2 — best in its comparison table at 3.7s inference (paper arXiv 2412.21037)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Paper arXiv 2412.21037 verified; benchmark table consistent across mirrors; Primary sources are community mirrors — official project page not found, officialUrl left empty; Stability AI Community License (non-commercial) caveat captured.