Seedance 2.0 Tested on Robot Tasks
DJiafei · x · 2026-07-18
The author tested Seedance 2.0 on the same robotic task, costing $7 to generate.
Their conclusion: motion appears more realistic, and scene consistency has improved, though the overall difference isn't massive. The biggest shortcoming remains the mechanical hand/gripper, which exhibits strange repetitive opening and closing. The author speculates this is because such grippers are rare in online Franka datasets, causing the model to encounter an OOD (out-of-distribution) scenario.
Finally, they pose the question: is this performance "good enough" to be used for robot training?
More from Embodied
- NVIDIA open-sources FLORA, a robotics data engine ~125x cheaper than GPT-6 Astra — nvidia · 2026-10-27
- Embodied AI startups need a different operational discipline, investor says — ___Mufasaa · 2026-10-27
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- GPT-6 Astra controls dexterous hands zero-shot at just 5.2% success in 500 trials — stingning · 2026-10-27
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27
- DexPort: a handheld interface with a real robot hand for dexterous-hand data collection — pabbeel · 2026-10-27