Bernini (ByteDance)
Bernini is a video production and editing framework that combines a MLLM-based semantic planner with a DiT-based render engine. As part of the project, full pipeline models and Bernini-R rendering variants derived from the Wan diffusion model are offered. It supports video processing tasks such as frame, style transfer, subtitle or watermark removal.
bytedance · Video & Media Production
Common commercial products in this category: Adobe Premiere Pro, Vimeo, Final Cut Pro
- You perform video editing and production with MLLM based semantic planner
- You perform video processing operations such as style transfer, subtitle and watermark deletion
- Using Bernini-R models, you run video render and inference commands
- • No published release
- • NeutralGitHub Issue
Can be installed in windows 11 + RTX 6000 pro 98GBVRAM?
“hi, can we install in windows and rtx 6000 will run it?”
- • NeutralGitHub Issue
Where is the MLLM planner?
“The paper presents Bernini as a two-part system: an MLLM-based semantic planner + a DiT-based renderer. Please clarify if/when the MLLM component will actually be published.”
My note
Alternative pages
More in: Video & Media Production
- supervision★ 51,438 · today
- Video.js★ 40,050 · today
- Video Use (Browser Use)★ 30,200 · 22 days ago
- SRS★ 29,401 · today
- memvid★ 17,322 · 3 mo ago
- filmcraft★ 16,978 · today