AI clipping studio · runs on your machine
Klipper watches your whole video, finds the moments worth posting, and cuts them into captioned vertical clips — transcribed locally by Whisper, scored by Claude, rendered by ffmpeg.
Real output style: 9:16, captioned, scored and ranked.
One pipeline, three stages, no editing timeline in sight.
Paste a link from YouTube, Twitch, TikTok, X and 1000+ sites — or drag in a file up to 2 GB. Got a 20-hour VOD? Livestream mode grabs just the best section.
Whisper transcribes the audio on your machine, word by word. Claude picks the moments with hook potential — or free mode finds them by audio energy alone.
Separate clips or one montage, in any aspect from 9:16 to 16:9, up to 4K, with captions, your @handle and a ducked music bed if you want them.
3 free videos when you sign up. Drop a video and go.
What to make
Output format
4K/8K only add real sharpness when the source is that high-res — otherwise they just upscale into much bigger files.
Give a timestamp to grab that exact spot, or leave it blank to auto-find the most exciting stretch by sampling the audio across the whole VOD.
The parts of clipping that eat your evenings, handled in one pass.
Karaoke-style captions burned into every clip, with the spoken word highlighted. Or flip the toggle off for clean footage.
Get separate clips, one stitched highlight reel at the length you choose, or just reframe the whole video — your call.
Point it at a multi-hour VOD. It samples the timeline by audio energy and grabs only the hottest section in HD — never the whole recording.
9:16, 4:5, 1:1 or 16:9 at 360p up to 4K. Reframe-and-fill or keep the whole frame over a blurred background.
Claude reads the transcript and scores every moment — or switch to free mode and let audio energy find the peaks. Sports and music work too.
Stamp your @handle on every clip and lay a music bed under the voice — auto-ducked so nobody gets drowned out.
Every new account gets 3 free videos (local engine, 720p, small watermark). 1 credit = 1 AI-clipped video — full quality, no watermark. Credits never expire.
Paystack isn't configured yet (set PAYSTACK_SECRET_KEY in .env.local) — checkout is disabled until then.
The app is free to run. You bring your own Anthropic API key, and AI picking costs pennies per video. Free mode (audio-energy picking), transcription and rendering cost nothing — they run on your machine.
Nowhere. Downloading, transcription and rendering all happen locally. The only thing sent over the network is the transcript text Claude reads to pick highlights — and nothing at all in free mode.
Yes — that's what the Livestream / heavy tab is for. It probes short audio samples across the VOD, finds the most exciting stretch (or jumps to your timestamp), and downloads only that section in HD.
Node 18+, an Anthropic API key for AI picking (optional), and about 2 GB of free RAM. No Python, no system ffmpeg — the binaries ship with the app.