Skip to content
Private beta · free · invite-only

Four hours live.
Thirty seconds worth posting.

Drop in a Twitch or Kick VOD, or upload your own file. Audience Labs finds the moments worth clipping and hands them back as vertical shorts with captions burned in.

  • Twitch VOD
  • Kick VOD
  • File upload
  • TikTok · Shorts · Reels

Thumbnails: clip footage from NASA, public domain. NASA does not endorse Audience Labs. Sources

How it works

One moment, from stream to clip.

Here is the liftoff at 2:14:40 from the stream above, step by step. Every clip goes through the same four steps.

  1. 01

    Transcribe

    Every word gets a timestamp, and laughs, shouts and sudden silences are marked too.

  2. 02

    Find the moment

    Claude reads the whole stream and picks the beats that stand on their own: setup, payoff and reaction.

  3. 03

    Cut on the words

    Start and end snap to the first and last spoken word. Timing comes from the transcript, never guessed.

  4. 04

    Render 9:16

    Reframed to vertical, captioned word by word, and encoded for TikTok, Shorts or Reels.

The editor's part

Claude reads the stream. The pipeline checks its work.

Audience Labs uses Claude, an AI model developed by Anthropic, for the part a human editor spends hours on: reading the stream, deciding which moments stand on their own, and explaining why.

  1. 01

    Reads the entire stream

    Long VODs are split into chapters first, so a story that crosses a break is not cut in half.

  2. 02

    Picks complete beats

    Setup, payoff and reaction together, not a one-liner without context.

  3. 03

    Grounded in the transcript

    Claude quotes the moment; the quote is found in the word-level transcript and the timing comes from there. If it can't be found, the clip is dropped.

  4. 04

    Grades it like an editor

    Separate Hook, Flow, Value and Trend grades, each with a written reason.

  5. 05

    Writes the packaging

    An on-screen hook line, plus a caption, hashtags and a title per platform. All editable.

Illustrative example, same moment as above. The model's guessed start is replaced by the transcript match.
Caption styles

Captions burned in, word by word.

Four styles, highlighted as each word is spoken. Pick one per clip or set a default.

  • Pod

    Clean, two words at a time

  • Reels

    Classic boxed subtitles

  • Beasty

    One loud word at a time

  • Hype

    Kinetic, word-by-word spotlight

Try it on a clip

The same options you get on every clip in the beta.

Size
Auto-hook line
Word-level edits

Burned in. Captions are part of the MP4, so they look the same on TikTok, Shorts and Reels.

Illustrative example.

Rendered by the real caption pipeline on public-domain NASA footage. NASA does not endorse Audience Labs. Sources

In the beta

Everything between the VOD and the upload.

  • Twitch, Kick or upload

    Link a VOD or upload your own recording, up to four hours.

  • Review queue

    Every clip waits for your approval. Export what you keep as a ZIP.

  • Re-render with edits

    Mis-heard a word? Fix it in the clip, change the style, render again.

  • Auto-hook line

    An optional title at the top for the first five seconds. You can rewrite it.

  • Platform-aware encode

    Bitrate and length limits for TikTok, YouTube Shorts and Instagram Reels.

  • Automatic cleanup

    Source VODs are deleted after 30 days unless you save them. Replaced clips stay 7 days.

FAQ

Questions, answered.

Who is Audience Labs for?

Streamers and the editors who work with them: anyone sitting on hours of Twitch or Kick footage who wants short clips without scrubbing through the whole thing.

Can I use it today?

Audience Labs is in a private, invite-only beta. Request access and you'll hear back when a seat opens.

What does it cost?

It is free during the private beta. Pricing for later hasn't been set, and beta testers will hear about it first.

Does it post to my accounts?

No. Clips are prepared for TikTok, YouTube Shorts and Instagram Reels and wait in a review queue. You download them and decide what goes out.

What happens to my footage?

Source videos are kept only to make clips and are deleted after 30 days unless you save them. You can delete clips at any time, and the privacy policy has the details.

Which languages work?

The spoken language is detected automatically. Today the analysis and captions are tuned for English and Turkish streams.

Who builds it?

One independent developer. If something breaks or you have an idea, the contact address in the footer reaches the developer directly.

Your next stream already has the clip in it.

Free during the private beta