Apple Researchers Unveil FastFrames Fast Autoregressive Video Diffusion Models
Start here
Apple researchers Gene Chou, Guandao Yang, Fujun Luan, Lars Mescheder, Wei Dong, Stephan R. Richter and Vladlen Koltun present FastFrames, a family of fast autoregressive video diffusion models that they say reaches interactive latency without sacrificing generation quality. According to the paper, the 1.3B variant runs at 36 ms per frame, or 28 responses per second, on a single B200 GPU while matching the quality of a state-of-the-art model that is nearly 8× slower. The 14B variant runs at 45 ms per frame, or 22 responses per second, on an 8× B200 GPU node and is said to deliver significantly better visual quality and realism. These figures and quality comparisons come from the researchers' own paper; independent verification is not reported.
Generated from reports · updated 27 min ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AIApple researchers present FastFrames, autoregressive video diffusion models reaching 28 RPS at interactive latency
FastFrames is a family of fast autoregressive video diffusion models that reaches interactive latency without sacrificing quality, per a paper by Gene Chou, Guandao Yang, Fujun Luan, Lars Mescheder, Wei Dong, Stephan R. Richter and Vladlen Koltun.
Heat for this event
Not enough continuous observations yet to draw a trend.