Transcribe Any Video for Free With This Local AI

Open Terminal. On Mac, press Cmd + Space, type “Terminal,” and press Enter.

First, install ffmpeg. This is the command-line video tool Whisper needs to read your files.

Mac: brew install ffmpeg

If you do not have Homebrew installed, ask Claude or ChatGPT how to set it up on your device.

Windows: choco install ffmpeg

Check that Python is installed. On Mac, type: python3 --version

If you see a version number, you are good. If not, ask Claude or ChatGPT how to install Python 3 on your device.

Install Whisper: pip3 install -U openai-whisper

That is it. You should only need to run these install commands once.

Pro tip: If pip3 is not recognized, try python3 -m pip install -U openai-whisper instead. If you want to double-check the install commands for your operating system, the Whisper GitHub page has everything.

Grab a video file you want to transcribe. Right-click it in Finder and copy the file path. You will need that for your Whisper command.

Run this in Terminal: python3 -m whisper "[your file path]" --model base

For example, if your video is in Downloads: python3 -m whisper "/Users/you/Downloads/my-video.mp4" --model base

Whisper will detect the language automatically, then pull the words and timestamps into your terminal. A gigabyte-sized, seven-minute video takes about two minutes to transcribe on a MacBook Pro.

• .txt for a plain text transcript
• .srt for closed captions, which you can upload to YouTube or import into your editor
• .vtt, .tsv, and .json for other formats

To control where the files go, add --output_dir and --output_format: python3 -m whisper "[your file]" --model base --output_dir "/Users/you/Downloads" --output_format txt

This guide uses the base model. There are several model sizes, and the GitHub page shows how much RAM each needs. You can go up to turbo, but base is the sweet spot for speed and accuracy.

Pro tip: The .srt file has timestamps baked in and imports directly into Premiere, DaVinci Resolve, Final Cut, and most other editors. If you are using the transcript for captions, use the .srt instead of the .txt.

Whisper can translate most non-English audio. Run the same command as before on a video or audio file and add the --task translate flag to the end.

python3 -m whisper "[your file path]" --model base --task translate

Without the --task translate flag, Whisper transcribes in the original language. With it, you get an English translation.

If you have a folder of recordings to process, you can batch the whole thing in one line. Open Terminal, cd into the folder, and run:

for f in *.mp4; do python3 -m whisper "$f" --model base; done

Every file in the folder gets transcribed. Walk away, then come back to a folder full of transcripts.