AI NewsOpen sourceAnnouncement
PeakCut is an open-source AI clip tool for videos and live streams
PeakCut is a self-hosted MIT tool that turns long videos and live streams into captioned vertical clips. You bring your own AI key.
Image: GitHub
Why it mattersA creator paying a monthly fee to a hosted clip tool can run PeakCut on their own server. The live-stream mode clips while the stream is on.
A creator who records on Twitch or publishes to YouTube wants short vertical clips with captions. They want them ready to post. Paid tools for that job charge a monthly fee for every minute of video they process. PeakCut is a new open-source alternative that runs on your own machine or server. The README calls it a way to turn long videos and live streams into captioned vertical clips. The code is on GitHub at haziqahmedqasmi/PeakCut under MIT. It was created on 14 November 2026. It had 393 stars the day after.
What it does, in the author's words
Paste a YouTube, Twitch or Kick link, or upload a file, the README says. PeakCut picks the moments and cuts them to a 9:16 vertical frame. It adds word-by-word captions. It then lets you post them straight to YouTube, Instagram and TikTok in one click.
Three caption styles are available. The real new part is the live-stream mode. The README says PeakCut can record a stream into a rolling buffer and clip as it happens. The signals it uses are loudness, chat activity and the transcript. The clip is ready while the stream is still on.
Bring your own AI key
The README lists the model providers it supports. The list is Gemini, OpenAI, Grok, Groq, OpenRouter, Claude and a local model through Ollama. It also works with anything that speaks the OpenAI chat format. You pay the provider for your own calls. Speech-to-text runs on your own machine with no third-party service.
The AI picks the interesting moments and writes the metadata. It does not generate the video itself. Multi-creator workspaces are built in. One set of keys and settings is held per creator. Each creator gets an activity log. A team using one server can see what every seat has posted.
Running it yourself
You need Python 3.10 or newer and ffmpeg, the README says. For a server, PeakCut has one-command deployment with PostgreSQL and automatic HTTPS through Docker Compose. There is no storage limit unless you set one.
Paid tools in this space charge per month or per processed minute. They also hold your footage on their own cloud. PeakCut keeps the files and the API keys on the machine you run. A creator who hits a paid tool's cap can read the source. They can change the moment-picking rules and run it on the server they already have.
Source
The repository is at haziqahmedqasmi/PeakCut under the MIT licence.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
