VibeClip is a new open-source, self-hosted tool that turns long videos into vertical, captioned shorts you direct by chatting. Upload a podcast, interview, or stream, and an AI agent finds the strong moments, reframes them to 9:16, burns in synced captions, and takes plain-language edits like "make clip 2 punchier" or "add a zoom at 0:05." It launched on Hacker News this week under the AGPL-3.0 license.
Try It: Turn One Podcast Into a Week of Shorts
Clone the repo, add your own LLM key, and drop in a long recording. VibeClip scores moments, cuts jumpcuts, reframes to vertical, and burns captions automatically, then lets you refine each clip by chat with single-command undo. Built-in styles include hormozi, mrbeast, podcast_minimal, and kinetic, so you can match a look without hand-editing. Because it runs locally with Docker, your footage never leaves your machine.
Why It Matters for Creators
Repurposing long content into shorts is the most time-consuming part of most creator pipelines, and the hosted tools that automate it charge monthly and process your video on their servers. A self-hosted, chat-driven alternative gives solo creators and small teams the same clip-finding and captioning workflow with full data control and no per-export fees. The chat interface also lowers the skill floor: you describe the edit instead of learning a timeline.
Key Details
Stack: Python with a web front end, transcription via faster-whisper running locally.
Bring your own model: Works with OpenAI, DeepSeek, and Claude-compatible endpoints for moment-scoring and edits.
License: GNU AGPL-3.0, fully self-hosted with Docker support and no data proxying.
Editing: Automated jumpcuts, reframing, caption burning, and music or sound effects, all refinable by chat.
What to Do Next
If you publish shorts from long-form content, spin up VibeClip in Docker and run one episode through it this weekend. Pair it with a browser transcription tool like the one in our Scribix coverage if you also need standalone transcripts. As an early-stage project at 10 commits, treat it as a workflow to bookmark and test rather than a finished product to depend on.
