IN THE CAN
Reel 1
The in-browser clip finder: loudness map, top 3 moments, 9:16 crop, Whisper auto-captions, title burn-in, WebM export, the live $CLIP ticker and the Premiere launchpad.
$CLIP · a meme coin for clippers
Drop a recording in. Cliprush listens to the whole thing, marks the three loudest moments on the timeline and cuts each one to a vertical 9:16 clip, with Whisper captions. All of it runs in your browser. Your file never gets uploaded.
Stays on your device. MP4 or WebM with an audio track work best.
English speech to text with word timing. The model (~45 MB) downloads once, then it's cached. Your audio stays on your device.
Somebody streams for four hours. Somebody else scrubs through all four to find the forty seconds worth posting. Most of the time the good part is easy to spot, because it's loud: the chat-wide scream, the crowd, the "NO WAY".
$CLIP is a meme coin for the people doing that scrubbing. The clip finder above is a free tool that does the first pass for you, on your own machine. A hosted auto-clipper that watches live streams is a concept we talk about, not a product that exists.
Your browser pulls the audio track out of the file with the Web Audio API, downsampled so even long VODs fit in memory.
Loudness is measured every tenth of a second, then compared against the last twenty seconds, so sudden jumps count more than steady noise.
The three highest-scoring windows that don't overlap become hype moments, each with a little lead-in before the peak.
The frame is cropped to 9:16 on a canvas, your title is burned in, and MediaRecorder writes a WebM you can save.
Loud doesn't always mean good. It's a first pass that saves scrubbing, not an editor's eye.
Launch a pump.fun coin from your own wallet. When it confirms, it goes on the Premiered here reel and rolls across the editor's ticker.
Coins launched from this page, newest first, with market caps from DexScreener. A list, not a recommendation.
IN THE CAN
The in-browser clip finder: loudness map, top 3 moments, 9:16 crop, Whisper auto-captions, title burn-in, WebM export, the live $CLIP ticker and the Premiere launchpad.
NEXT
$CLIP launching on pump.fun. The contract address goes up here and on X only.
CONCEPT
Smarter scoring that mixes loudness with speech, scene cuts and motion. Ideas, no dates.
CONCEPT
A community clip wall where people post what the finder found. Not built.
No. The file is opened with a local object URL and read by your own browser. There is no server behind this page that could receive it.
Browsers can only decode the formats they support. MP4 (H.264 with AAC audio) and WebM are the safest. A video with no audio track can't be scored, because the score is loudness.
The audio is decoded at a low sample rate to save memory, so a couple of hours usually works on a laptop. Very large files (over about 1.5 GB) are refused to avoid crashing the tab.
WebM in most browsers (MP4 where the browser only records that). It records in real time, so a 15 second clip takes 15 seconds. Many platforms accept WebM, some need converting first.
Whisper tiny (English), an open speech-to-text model, runs inside your browser through transformers.js. The first time, the model files (~45 MB) download from Hugging Face and get cached. Your audio is never uploaded. Tiny models make mistakes, especially with music or shouting, so check before posting.
No. It's a meme coin. The tool is free for everyone, holder or not, and the coin has no expected return.