Get Started
← Creator toolkit · Mission 1 of 13

Subtitles for any video, free and offline

Big, bold, Reels-style captions burned into your video, made by OpenAI's Whisper on your own laptop. You also get an .srt for YouTube and a plain transcript. No subscription, no upload, no graphics card.

25 minBeginnerTools: FreeStep 1 free
Tested 7 Oct 2026 · whisper-ctranslate2 0.5.8 (faster-whisper 1.2.1), uv 0.11, FFmpeg 9.0.2, 16GB laptop with no graphics card, Windows 11
Short, punchy Reels-style captions, burned in big, bold and centred the way creator apps do it
A one-line command that turns any video or audio file into subtitles
An .srt file you can upload to YouTube or import into CapCut, plus a plain .txt transcript
  • Real speed and accuracy numbers from an ordinary laptop, so you know which model to pick
Before you start
  • 8GB of RAM is enough for the recommended model
  • About 750MB of free disk space (the app plus one model)
  • A video or audio file with speech in it (mp4, mov, mp3, m4a and wav all work)
  • Level 0 helps, but every command is here to copy
● What you'll have in 25 minutes
Preview
Captions made by Whisper on an ordinary laptop in 25 seconds, burned in with FFmpeg. Footage: President Kennedy at Rice University, 1962 (NASA, public domain).
01/06Install Whisper with one command0 done
STEP 01 of 06~5 min

Install Whisper with one command

Get a tool that runs OpenAI's speech-to-text model on your own CPU, with no Python fiddling.

  1. Whisper is OpenAI's free, open-source speech recognition model. We'll run it with whisper-ctranslate2, a small command-line app built on faster-whisper ↗. It gives the same results as OpenAI's original, runs well on a normal CPU, and can read video files directly.

  2. First install uv, a tool that installs Python apps cleanly in their own box. Open a terminal (PowerShell on Windows) and paste this (it's the official installer from astral.sh ↗):

    Windows
    powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
    Mac / Linux
    curl -LsSf https://astral.sh/uv/install.sh | sh

  3. Close the terminal and open a new one. Then install Whisper:

    Terminal
    uv tool install whisper-ctranslate2 --with "av<19"

  4. The --with "av<19" part keeps Whisper's audio reader on a version that works. A release in late September 2026 breaks faster-whisper with a metadata_errors error.

    Output
     + faster-whisper==1.2.1
 ...
 + whisper-ctranslate2==0.5.8
Installed 1 executable: whisper-ctranslate2

  5. uv downloads its own copy of Python if it needs one, so you don't have to install Python yourself. The whole app takes up about 243MB.

    Terminal
    whisper-ctranslate2 --version
    Output
    whisper-ctranslate2 0.5.8
✓
Checkpointwhisper-ctranslate2 --version prints a version number.
If it breaks'whisper-ctranslate2' is not recognised + 2 more
'whisper-ctranslate2' is not recognised
Close every terminal and open a new one. Still failing? Run uv tool update-shell, then open a new terminal again. It adds uv's app folder to your PATH.
Transcribing fails with an error mentioning metadata_errors
You installed without the version pin. Run uv tool install whisper-ctranslate2 --with "av<19" --force and try again.
'uv' is not recognised
The uv installer only takes effect in new terminal windows. Close the terminal, open a fresh one and try again.
Still stuck? Ask the community
Next mission · Creator toolkitShrink, trim and reformat any video in seconds
→