Seedance 2.5 tested with a first-person dog fetch scene to stress POV consistency
Celestial_69000 · reddit · 2026-09-07
A Reddit user tested Seedance 2.5 with a first-person dog fetch scene, checking how well the video model maintains POV framing, hand movements, and the dog's motion across the throw, catch, and return sequence — a grassroots stress test of the model's long-shot consistency.
More from Multimodal
- GPT-6 Astra geometry to Dreamina Seedance 2.5: a Blender-to-video pipeline — mhdfaran · 2026-09-08
- tldraw flash lets you create animations from scratch or remix ready-made templates — max__drake · 2026-09-08
- GPT-6 Astra builds an M.C. Escher 3D scene and renders a styled video in under an hour — bennash · 2026-09-08
- Creator shares unsettling AI video demo with intense camera movement, prompt included — umesh_ai · 2026-09-08
- Omni 1.1's video extension looks pretty seamless in early hands-on — fofrAI · 2026-09-08
- GPT-6 Astra + BlenderMCP beats Tripo and Meshy in hands-on 3D asset generation test — sidahuj · 2026-09-08