Flux 3 X Mimic: The Next Generation of Video-Action Models
Black Forest Labs says an early FLUX 3 backbone is now powering a robotics model tested and deployed at Audi.
FLUX-mimic adds action prediction to FLUX 3’s video/audio/image training rather than using a separate robotics foundation model. The company says video prediction dominates training cost and gives the model a usable representation of contact, motion, weight, and cause and effect. mimic decodes robot actions from FLUX features for factory tasks including kitting, ECU insertion, component assembly, and handling flexible seals and cables. BFL says the system reaches 101ms reaction times in mimic’s deployment stack. HN · Frontpage AI's note
FLUX-mimic adds action prediction to FLUX 3’s video/audio/image training rather than using a separate robotics foundation model. The company says video prediction dominates training cost and gives the model a usable representation of contact, motion, weight, and cause and effect. mimic decodes robot actions from FLUX features for factory tasks including kitting, ECU insertion, component assembly, and handling flexible seals and cables. BFL says the system reaches 101ms reaction times in mimic’s deployment stack. HN · Frontpage AI's note
score 6
Discussions
- hn · 258 points · 40 comments
- hn · 271 points · 44 comments
- hn · 280 points · 45 comments
- hn · 280 points · 45 comments
- hn · 284 points · 45 comments
- hn · 289 points · 47 comments
- hn · 292 points · 47 comments
- hn · 296 points · 47 comments
- hn · 297 points · 47 comments
- hn · 297 points · 47 comments
- hn · 302 points · 47 comments
- hn · 305 points · 48 comments
- hn · 306 points · 48 comments
- hn · 310 points · 48 comments
- hn · 310 points · 48 comments