GPU Whisper large-v3
Near-perfect accuracy on Ryzen/RTX. Fast enough for iterative editing.
Your videos, editable by text
Subtitle-driven auto video editor. GPU Whisper large-v3. Hermes MCP plugin. One-line install.
Pipeline
From video to edit in three steps
Transcribe, detect silence, render. Each stage is a standalone tool you can also use independently.
01
Transcribe
GPU Whisper large-v3
Automatic transcription with near-perfect accuracy. Fast enough for iterative editing.
02
Analyze
Smart Silence Detection
Detects pauses and sentence boundaries. Mark what to keep, cut the rest.
03
Render
ffmpeg + libx264
Renders the final cut at sentence boundaries. SRT/ASS subtitle support built in.
Built for Hermes Agents
Features
Everything you need to edit video entirely from text. No timeline, no dragging, no cloud.
Smart Silence Detection
Configurable gap threshold. Cut at sentence boundaries, not mid-word.
SRT/ASS Subtitle Support
Parse existing subtitles or generate new ones. Edit timing with precision.
MCP Plugin (8+ Tools)
13 FastMCP 3.0.2 tools: mark, cut, edit, pipeline. Integrates with any Hermes agent.
One-Line Install
npm-free. curl | bash or iwr | iex. Docker image also available.
Local, Private, Free
All processing on your GPU. No cloud dependencies. MIT licensed.
Built for speed
From development to deployment — one tool, one command.
CLI-first. No timeline, no dragging.
GPU-accelerated. Local inference.
Live Demo
Try it in your browser
Watch hermes-editing-yt process a real video — step by step. Pick a pipeline below.
hermes-cli — autoedit
$Processing...
Simulation — real outputs may vary by GPU, file size, and threshold config
New Release — v1.0.0
Built by one dev, for everyone
hermes-editing-yt is a fresh v1.0 release — built solo by Capslockb. No team, no funding, no VC. Just a tool I needed for myself.
Contribute
Found a bug? Want a feature? PRs welcome. The project is open source and actively maintained.
Contributing guide →Open Issues
Good first issues tagged for new contributors. Pick one up or suggest your own improvement.
Browse issues →Spread the word
Star the repo, share it with your team, or write about it. Every bit helps a solo dev project grow.
Star on GitHub →v1.0.0 · MIT · built by @Capslockb · first contributors welcome
See it in action
Three pipelines for different depths of automation.
automark
Plan only, no rendering. Transcribe + silence detect + produce edit marks.
hermes run --pipeline automark \ --input video.mp4autocut
Slice from existing SRT. Read subtitle timestamps and produce a cut video with no re-transcription.
hermes run --pipeline autocut \ --input video.mp4 \ --srt subtitles.srtautoedit
Full pipeline. Transcribe, detect silence, render final cut. One command, done.
hermes run --pipeline autoedit \ --input video.mp4 \ --output edited.mp4 \ --threshold 0.5Install in one command
Choose your platform and paste the command into any Hermes chat.
bash <(curl -sSL https://raw.githubusercontent.com/Capslockb/hermes-editing-yt/main/installer/install.sh)