Open-source video insertion model PISCO accepted to NeurIPS, needs just one keyframe
_vztu · x · 2026-10-01
PISCO (Precise Video Instance Insertion With Sparse Control), one of the few open-source video generative models from the US, has been accepted to NeurIPS. Sparse control lets it insert objects with temporal consistency across an entire video using only a few keyframes—even a single frame—addressing flicker, drift, and the 'pasted-on' look of naive compositing.
More from Multimodal
- Kling AI 4.0 taekwondo sparring clip shows natural multi-shot camera work — Aiden_Tech_Ai · 2026-10-01
- Utopai X, built on MiniMax H3, nears the top of Artificial Analysis text-to-video leaderboard — ArtificialAnlys · 2026-10-01
- Utopai X Use-Case Breakdown: First in Social Media Content, Second in Marketing Behind Wan 3.0 — ArtificialAnlys · 2026-10-01
- Utopai X, Post-Trained on MiniMax H3, Debuts at #2 on Artificial Analysis Text-to-Video Leaderboard — ArtificialAnlys · 2026-10-01
- Ideogram Demo Shows In-Image Text Editing and Translation — CodeByPoonam · 2026-10-01
- Two-model workflow makes a Dots ad in 7 hours, burning only 5% and 15% of quota — arrakis_ai · 2026-10-01