declare-lab/TangoFlux

[ICLR 2026] TangoFlux: Super Fast and Faithful Text to Audio Generation with Flow Matching

54
/ 100
Established

Uses Diffusion Transformers (DiT/MMDiT) conditioned on text and duration embeddings with rectified flow matching to learn trajectories in a VAE-compressed latent space. The three-stage training pipeline incorporates CRPO (Clap-Ranked Preference Optimization), which iteratively synthesizes preference pairs and applies DPO loss to align generated audio with human preferences. Integrates with Hugging Face (model hosting and accelerate training framework), ComfyUI for node-based workflows, and provides Python API, CLI, and web interface access.

843 stars.

No Package No Dependents
Maintenance 10 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 18 / 25

How are scores calculated?

Stars

843

Forks

76

Language

Jupyter Notebook

License

Last pushed

Jan 28, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/generative-ai/declare-lab/TangoFlux"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.