Netflix Releases VOID Video Inpainting Model
Netflix researchers on April 3, 2026 published a preprint and released VOID, a vision-language model that removes objects from video and inpaints physically plausible outcomes. The model, authored by Netflix and Sofia University researchers, is available on Hugging Face and was preferred in human tests—64.8 percent versus 18.4 percent for Runway—across synthetic and real-world scenarios. VOID targets film editing and automated video-manipulation workflows.
Key Points
- 1Introduces VOID, a vision-language model for object removal and physically-plausible video inpainting
- 2Shows higher human preference—64.8% versus 18.4% for Runway—on synthetic and real scenarios
- 3Enables filmmakers and editors to alter scenes without reshoots; raises manipulation and authenticity concerns
Scoring Rationale
VOID is a notable research release with runnable code on Hugging Face and strong human-preference results, boosting novelty and actionability. Score moderated because the work is a preprint (not peer reviewed) and coverage here is brief rather than deeply technical.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
