Back to list

PeakCut

Turn long videos and live streams into captioned vertical clips, ready to post — self-hosted, with your own AI key.

Callie Shaw
Author
@Callie Shaw
Project ShowcasePhoto & VideoOnlineclaude-code
0
0

Project intro

PeakCut turns long videos and live streams into short vertical clips that are ready to post. Paste a YouTube, Twitch or Kick link, or upload a file, and it transcribes the audio, asks an AI which moments would hold a stranger's attention, cuts them to 9:16, and adds word-by-word captions and a hook.

It also clips while the stream is still live: it records into a rolling buffer and cuts highlights as they happen, using loudness, chat activity and the transcript. Finished clips can go to YouTube, Instagram and TikTok from a queue, with spacing and daily caps, and each creator gets a workspace with channels, clip settings and an activity log.

Everything runs on your own machine or server — Python with FastAPI, ffmpeg and yt-dlp, with speech transcribed locally so audio never leaves the machine. It works with Gemini, OpenAI, Grok, Groq, OpenRouter, Claude, a local model through Ollama, or anything OpenAI-compatible, and with no key at all it still picks clips by loudness. It has 1,185 stars on GitHub with 404 forks, 24 commits, was created on 14 November 2026, and was last pushed on 17 November 2026.

Two honest limits: there is no public demo to try — it runs on your own computer at a local address, and the first start downloads a speech model of about 480 MB. It is also five days old at this check, with no releases published yet, so expect a fast-moving setup.

Actions

Visit project URL

Tech stack

pythonfastapiffmpegyt-dlp

Project info

Project category
Photo & Video
Project status
Online
Collaboration intent
None

Comments (0)

ME
0/1000
Loading comments...