Pyramid Flow
Open text-to-video that trains cheap — 10-second 768p clips via pyramidal flow matching.
About
Pyramid Flow, released October 10, 2024 by researchers at Peking University and Kuaishou, is an open text-to-video model built on a technique called pyramidal flow matching. By generating video as a coarse-to-fine pyramid, it trains far more cheaply than standard diffusion video models while producing 10-second clips at 768p and 24fps, plus image-to-video. The code is MIT and the weights are free for noncommercial use and for commercial users under $1M in annual revenue — not an OSI-approved open license.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Paper-reported: competitive quality with open video models at a fraction of training cost; user-study preference over Open-Sora and CogVideoX-2B (team-reported).
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: Released Oct 10, 2024 by PKU/Kuaishou researchers; Code MIT; weights carry a revenue-capped custom license; Research project — ongoing maintenance uncertain. License: Custom: free for noncommercial use and commercial use under $1M annual revenue; code MIT..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · Released Oct 10, 2024 by PKU/Kuaishou researchers
- · Code MIT; weights carry a revenue-capped custom license
- · Research project — ongoing maintenance uncertain