11333
Finds the moments in a long stream video where the streamer farts, and lets you save them as clips. Everything happens on your phone; your video is never uploaded.
💡 You can drag video files (or a 11333 backup .json) from a folder and drop them anywhere on this page.
1. Pick your video (or several)
2. Options (optional)
Results
🎬 Auto-comp
Scanning…
👁 Peek at included clips (optional)
Remove any you don’t want before compiling. Closed by default — you don’t need this.
📋 Show all lists and settings
Tap ▶ to watch a moment. “Save clip” downloads just that bit.
⌨ In the review player: K keep · X or R reject · N 👊 knock · 1–6 👊 🤧 👃 💋 🪀 ❓ what the sound was · ← → previous / next · Space play / pause · [ start 2 s earlier · ] end 2 s later
🎬 Make a compilation
Joins the clips you’ve kept into one video, on your phone.
Only kept clips go in: the ones you passed with ✓ Keep, plus ⚡ auto-kept ones (scored at or above your auto-keep score, if that’s switched on). Clips you haven’t reviewed yet are left out.
↕ Change the order
With none of these ticked, clips are joined without re-encoding: fast and full quality. Ticking any of them re-encodes the video, which is slower (roughly the length of the compilation or more).
⚙️ Settings
Changes here update the results instantly. No need to scan again.
🧑🎤 Streamers & learning
Every ✓ Keep / ✗ Reject in the review player teaches the app what this streamer's farts (and false alarms) sound like. Each streamer has their own examples; new streamers start from everyone's examples.
One automatic backup file is saved each week, into Downloads/11333. This phone cannot replace a file, so after two unconfirmed files the app waits while you delete the oldest. The file keeps your choices, every learned example and the background average.
❓ How it works & tips
- Listen: a sound-recognition model (Google’s YAMNet, which knows a “Fart” sound) listens to the whole video. A second check looks for sudden low rumbly bursts.
- Reactions: laughing or giggling just after.
- Look: around each suspicious sound it watches the streamer’s body for a lean or hip lift, and the camera for a sudden swing.
- Learns: your Keep/Reject choices train a small personal model per streamer, so similar sounds score higher next time and known false alarms are pushed down.
- Each clue adds to a confidence score. Moments above your sensitivity become hits.
Speed: the sound is checked on several processor cores at once, and the AI only listens closely to suspicious moments, so an hour of video takes roughly 1–4 minutes on a recent phone (more with thorough or deep scan). Results are saved, so opening the same video again is instant.
Tips: if you get too many false alarms, lower Sensitivity or switch off “Low rumble burst”. If the streamer is a small webcam over a game, pick the matching corner above. Saved clips go to your Downloads folder.
📊 Stats
Counted on this device from your saved scans; nothing is sent anywhere. False alarms = clips you rejected (incl. 👊 knocks) out of all the clips you decided yourself.
⏱ Scan times
How long each part of your recent scans took on this device (seconds). Saved with each scan from v2.30, so they are also in backups and training packs.