Search by word
Look up a name, number, or phrase in the transcript to go straight to that spot in the video.
Theme
Video editing with a timed transcript
When every line of a transcript has an exact timestamp, you can search a video like a document.
Live demo
Runs in your browser.
Find the sentence you need, grab its start and end times, and make the cut in an editor yourself or with help from an AI assistant.
Scrubbing through video takes time, and it is easy to lose your place. Reading text is much faster. When every line carries a start and end time, you can find what you need on the page and go straight to that second in the footage.
Look up a name, number, or phrase in the transcript to go straight to that spot in the video.
You can see filler words, false starts, repeated takes, and tangents on the page much faster than by scrubbing the timeline.
Every line has a start and end time, so you can cut right where the sentence starts and stops instead of guessing a beat early or late.
KolWrite generates the transcript and keeps the timing intact. You decide what stays in the cut, and you can have an AI assistant suggest which parts to keep.
Ask the built-in AI chat where a topic came up or have it list the main points, then jump directly to those timestamps.
Export your transcript as SRT, TXT, or JSON and paste it into an AI assistant. Ask it to pick out which segments to keep along with their start and end times.
Use the suggestions as a checklist to go through. Listen to each spot before making a cut, since an assistant can misread tone and context.
Upload your file to KolWrite to get a timestamped transcript. If more than one person is speaking, turn on speaker labels.
Read through the text, run a search, or ask the AI chat. You can also paste the exported transcript into another AI assistant and ask for the start and end times of the sections worth keeping.
Open the video in the KolWrite editor, make your cuts on the timeline using those timestamps, and export. Those same timestamps will work in any other video editor too.
Burn captions directly into the video, or export an SRT file to use in another program. Hebrew and Arabic captions are properly formatted right to left.
You can also run everything KolWrite does with a recording through our API. Send the address of an audio or video file from your own code or an AI coding agent, and you'll get back the transcript with word timings, speaker labels, and an SRT file if you request one.
The JSON output gives you the start and end time for every segment, plus individual word timings on models that support alignment. Turn on speaker separation to include a speaker label on each segment.
Pass output_format as srt if you want subtitles directly from the API. If you are using the Oracle models, you can also set translate_to to get the text back in another language.
The API console publishes plain-text instructions for an agent to read before writing code. This includes agent-instructions.md, along with llms.txt, which links to the full reference.
Example request. The key comes from an environment variable, and the console has the full reference.
# Times come from the transcript (seconds). Keep 12.4 s to 18.9 s of the source:
ffmpeg -i input.mp4 -ss 12.4 -to 18.9 -c:v libx264 -c:a aac part1.mp4
# Burn the captions into the picture:
ffmpeg -i input.mp4 -vf subtitles=captions.srt captioned.mp4
First, get a transcript with start and end times on every line. Choose the parts you want to keep on your own, through the AI chat in KolWrite, or by having an assistant pull segments from the text. Next, make your cuts at those timestamps in a video editor and play back each one to check it.
A text-only assistant cannot cut video files on its own. Instead, it reads a timed transcript and gives you the timestamps, while a tool like the KolWrite video editor or ffmpeg handles the actual cutting. Developers can link the two using the KolWrite API.
Use SRT or JSON for cutting, since the assistant needs the timestamps. SRT is compact and includes the times, while JSON has the timing for every single word. TXT is the smallest file.
Yes. A timed transcript shows you where pauses and filler words happen so you can cut them out. Just be sure to listen back to every cut, because trimming too close to the words can make speech sound unnatural and clipped.
Yes. You can transcribe Hebrew or Arabic audio, edit the right-to-left text and captions, and cut the video using the same timestamps. Make sure to preview burned-in captions before you publish.
Yes. The KolWrite API outputs segments with start and end times along with word-level timings. You can write a script or use an AI coding agent to select the segments and run ffmpeg to cut the video. Check the developer section below and the API console for details.
KolWrite's transcription is also built into KolHive, an AI OS with an assistant that can work on the video timeline. Check kolhive.com to see what is available right now.
Upload a video that has a few retakes. Read through the transcript, mark the takes you want to keep, and check the timestamps against the footage.