Free tools now convert any file in your browser. Video, audio, images and PDF, nothing uploaded.
Free AI tool
Turn a recorded transcript into a clean training script.
0 / 24,000
Worked example
Based on a short recording transcript written for this page. Every fact in it comes from 29 CFR 1910.1200, Hazard Communication, US Electronic Code of Federal Regulations.
11 filler words removed. Cut the opening apology, the aside about losing the room, the correction about the telephone number, and the closing remark about next time.
Six things on every label
Every shipped container carries six pieces of information. Once you know the six, a label stops being decoration and starts being something you can read at a glance.
Visual: A container label drawn as six numbered callouts around a diamond pictogram.
Identifier, then signal word
First the product identifier, which is what the substance actually is. Then the signal word, and there are only two of those: danger, or warning.
Visual: Two label fields highlighted in sequence, with the two signal words shown as the only options.
What it does to you
Then the hazard statements, which say what it does to you, and the pictograms, the red bordered diamonds that carry the same message without words.
Visual: A hazard statement line next to the matching pictogram, joined by a short connector.
What to do, and who to call
Then the precautionary statements, what to do and what not to do. And last, the name, address and telephone number of the party responsible for the chemical.
Visual: A precaution list on the left, a responsible party block on the right with a phone icon.
16 sections, reachable on shift
Safety data sheets sit behind the label. Sixteen sections, always in the same order, and they have to be reachable in the work area on every shift. A locked office does not count.
Visual: A grid of sixteen numbered tiles beside a work area floor plan, with the office marked as not acceptable.
1. Paste the transcript. 2. We count the filler. 3. Get the clean script. 4. Record or render.
What Interlace does
Real Interlace output, unedited. Shown silent and sped up.
Your transcript
“…so, um, the main change this quarter is how customer data moves between, you know, our systems…”
Questions
In your browser, not by the model. We count a fixed list of filler words in the transcript you pasted, so the number on screen is something you could count yourself rather than something a model asserted.
That is the whole job. It is told to keep every fact stated and to add none, and to cut only filler, repetition, small talk and remarks about the recording itself. Read the result against the transcript before you record it.
Loom, Zoom and YouTube all export one. Paste the plain text. Timestamps and speaker labels do no harm, though the cleaner the paste, the cleaner the script.
24,000 characters, which is roughly ten pages of dense text. The counter on the page tells you when you are over, and anything past the cap is ignored. For something longer, run it a section at a time.
To OpenAI, which generates the result. The text you paste, or the text read from your file, goes from our server to OpenAI's API and nowhere else. OpenAI does not use API data to train its models, and we do not log or keep your input. The result is cached for up to 30 days, so the same text run again comes back fast.
No. The tool is free and there is no sign in. You only need an account when you want Interlace to turn the result into a finished video.
Create an account and start from a document, script, or prompt.