EditorZero
Self-hosted, transcript-driven editing suite that keeps raw footage off vendor clouds.
SPECIFICATIONS
- STATUS: [PILOT]
- YEAR: 2025
- TYPE: AI Post-Production
- STACK: Python, Llama 3.1 8B, Scene Detection, Transcript UI, Self-hosted GPU
Overview
EditorZero bundles a Python backend with Llama 3.1 8B so editors can cut directly inside the transcript while keeping footage on their own metal. It recreates the Descript experience without surrendering terabytes of source files to a SaaS. The sync engine keeps clips, gaps, and speakers locked to the text editor while the AI proposes trims, highlight reels, and subtitle passes.
Signature Moves
- Sync Engine: Text edits ripple through the multitrack timeline, preserving frame accuracy even during aggressive cuts.
- Hybrid Intelligence: Prompt libraries pair Llama outputs with classical scene detection to auto-build rough cuts, remove dead air, and flag standout quotes.
- Local Rendering: Subtitles, transitions, and LUTs render on the same server that stores media, so every export stays private and consistent.
Stack Notes
- Python orchestrates whisper-style alignment, numpy driven waveform analysis, and GPU acceleration for renders.
- The transcript UI is a lightweight web app that mirrors edits to the server through websockets.
- Deployment script ships as a Docker compose file for quick installs on studio NAS hardware.
Outcomes
Studios now run Descript-caliber tooling without recurring licenses, moving sensitive footage through an air-gapped workflow. It is the preferred option when the footage involves NDAs, unreleased IP, or simply demands sovereign control.