# Toward More Controllable AI Video Editing: An Early Research Exploration at Netflix

[Netflix](https://yomu.fyi/company/netflix) · Netflix Technology Blog · Jun 23, 2026

**Type:** Announcement

## Summary

Generative video editing tools often regenerate entire video clips, which inadvertently modifies untouched scene details or violates physical continuity during object deletion. To provide artists with precise control, two research prototypes were developed: Vera and VOID. Vera uses a layered video diffusion framework with a Mixture-of-Transformers architecture to jointly generate separate edit and alpha matte layers, compositing them with original footage while leaving unchanged pixels untouched. VOID addresses object deletion by conditioning on interaction-aware quadmasks and using a second-pass appearance refiner to reconstruct scenes with plausible physical continuity. In a user study evaluated across 75 real-world scenarios, reviewers selected VOID 64.8% of the time over six baselines.

## Context

Generative video editing models frequently regenerate every pixel, inadvertently altering preserved identities, performances, and scene details, or ignoring physical continuity when erasing objects.

## Approach / What changed

Netflix introduced two models: Vera, a layered video diffusion model using a Mixture-of-Transformers architecture with three separate DiTs for edit, alpha, and composite outputs; and VOID, a video inpainting model fine-tuned on the CogVideoX-Fun-V1.5-5b-InP backbone using interaction-aware quadmask conditioning and a second-pass video appearance refiner.

## Takeaways

- Vera relies on a Mixture-of-Transformers design with three separate DiTs that maintain independent QKV and FFN weights while sharing joint self-attention across output tokens.
- VOID incorporates quadmask conditioning and a second-pass appearance refiner to remove objects while maintaining physically plausible downstream scene dynamics.
- In a user study against six baseline models across 75 real-world scenarios, VOID was selected as the most realistic output 64.8% of the time.

**Tags:** [Machine Learning](https://yomu.fyi/topic/machine-learning)

[Read original post](https://netflixtechblog.com/toward-more-controllable-ai-video-editing-an-early-research-exploration-at-netflix-eb8160ed60a2)