Adobe's MoviEdit Tackles Long-Form Multi-Shot Video Editing With 87K Auto-Generated Pairs
adobe · hf · 2026-11-04
Adobe and KAIST introduce MoviEdit, an end-to-end framework for long-form multi-shot video editing that maintains coherence across scene, viewpoint, and appearance changes.
Highlights:
- Data: automatic pipelines spanning synchronized multi-view footage, real multi-shot videos, and generated content yield 87K paired editing videos without manual annotation
- Architecture: cascaded bidirectional-autoregressive design — a lightweight bidirectional stage drafts a globally coherent edit on compressed representations, then autoregressively refines it across the full video
- Benchmark: introduces MoviEdit-Bench for systematic evaluation
- Results: substantially outperforms open-source models and matches industry frontier models on text- and image-guided editing
Project page: multi-shot-edit.github.io
More from Multimodal
- MiniMax H3 video gen can't keep characters dead: Reddit user's RTX 3090 workflow test — DirectSentence9823 · 2026-11-04
- Bullish take: Spawn 7 will convince everyone the 3D internet era has begun — TAbrodi · 2026-11-04
- AI video made with Seedance 2.5 on Higgsfield hailed as a 'masterpiece' — SimplyAnnisa · 2026-11-04
- Prompting "photorealistic" makes Flux images look more like 3D renders, Reddit test shows — Confident_Ad2351 · 2026-11-04
- Pika Releases H3-Flex-Turbo LoRA for 2/4/8-Step Image-to-Video Generation — Pika-Labs · 2026-11-04
- Nanjing University's SeMo3D Unifies 3D Generation, Editing, and Completion in One Model — nanjinguniv · 2026-11-04