What WhisperX does and why you might want it

WhisperX is free software that automatically creates captions for audio and video files on your computer. It listens to what's being said, writes it down as text, and syncs those words to the exact moment they're spoken. You can use it on recordings you already have — podcasts, interviews, home videos, lectures — without uploading them anywhere or paying a service.

The main reason to install it yourself instead of using a web-based captioning tool is privacy. Your files stay on your own computer. Nothing gets sent to a company's server, so there's no record of what you captioned or when. If you're working with sensitive audio — medical appointments, therapy sessions, confidential meetings — that matters.

WhisperX runs on Windows, Mac, and Linux. The installation takes about 15 minutes if you follow the steps in order. You'll need a computer with at least 4 GB of RAM and about 3 GB of free disk space for the software itself, plus room for your video files.

Key Takeaways

  • WhisperX requires Python to run, which you install first, then WhisperX itself through a command-line tool called pip.
  • The software is free and keeps your audio files private because everything happens on your own computer, not on someone else's server.
  • After installation, you run WhisperX by typing a command that points to your audio or video file, and it creates a caption file you can use in video players or editing software.
  • If the command-line steps feel unfamiliar, desktop applications like Faster-Whisper-Transcription provide the same core tool with a clickable interface instead.

Install Python first

WhisperX needs Python — a programming language — to run. You probably don't have it on your computer yet. Go to python.org, click the yellow "read Python" button, and choose the version for your operating system (Windows, macOS, or Linux). read the latest stable version (3.10 or newer).

Run the installer. On Windows, check the box that says "Add Python to PATH" before you click Install — this step matters, because it lets your computer find Python from anywhere. On Mac, the installer handles this automatically. Let the installation finish completely.

To check that Python installed correctly, open a command prompt (Windows) or terminal (Mac/Linux). Type python --version and press Enter. You should see a version number like "Python 3.11.5". If you see "command not found" or "not recognized", Python didn't install properly — try the installer again and make sure you checked the PATH box on Windows.

Install WhisperX using pip

Now that Python is on your computer, you can install WhisperX. Open your command prompt or terminal again and type this line exactly:

pip install openai-whisper

Press Enter and wait. Pip (Python's package installer) will read WhisperX and everything it needs. This can take a few minutes depending on your internet speed. You'll see text scrolling past — that's normal. When it finishes, you'll see a line that says "Successfully installed" with a list of packages.

Next, install one more piece that makes WhisperX faster and more accurate. Type:

pip install git+https://github.com/m-bain/whisperx.git

Press Enter and wait again. This downloads the WhisperX code itself from GitHub (a code-sharing website). When it finishes, you're ready to use it.

Run WhisperX on your first audio file

Put an audio or video file somewhere you can find it easily — your Desktop or Documents folder works well. WhisperX works with MP3, MP4, WAV, M4A, and most other common formats.

Open your command prompt or terminal again. You need to navigate to the folder where your file is. On Windows, type cd Desktop (if your file is on the Desktop). On Mac or Linux, type cd ~/Desktop. Press Enter.

Now type the command to run WhisperX. Replace "yourfile.mp4" with the actual name of your file:

whisperx yourfile.mp4 --model base --language en

Press Enter. WhisperX will start working. The first time you run it, it downloads a language model (about 140 MB) — this takes a minute or two. After that, it processes your audio. A 10-minute file usually takes 2 to 5 minutes to caption, depending on your computer's speed.

When it finishes, you'll see a new file in the same folder with a name like "yourfile.vtt" or "yourfile.srt". These are caption files. You can open them in a text editor to read them, or load them into a video player or editing software to see captions synced to your video.

Understanding the command options

The command above uses three options: --model base, --language en, and the filename. You can change these depending on what you need.

The model controls accuracy versus speed. "base" is a good middle ground — it's reasonably fast and reasonably accurate. If you want faster results on a slower computer, use --model tiny. If you want higher accuracy and don't mind waiting longer, use --model small or --model medium. Larger models need more RAM and disk space.

The language option tells WhisperX what language to listen for. --language en is English. If your audio is in Spanish, use --language es. French is fr, German is de, and so on. If you leave this out, WhisperX tries to guess, which sometimes works and sometimes doesn't.

You can also add --output_format srt to get a different caption format, or --output_dir /path/to/folder to save captions somewhere other than the same folder as your file. For most uses, the basic command above is enough.

Using captions in video software

Once you have a caption file, you can load it into almost any video player or editing software. In VLC (a free video player), open your video, then go to Subtitle → Load Subtitle File and choose your .vtt or .srt file. In DaVinci Resolve (free video editing software), import your video, then right-click the clip and choose "Import Subtitle File".

If you're uploading video to YouTube, you can upload the caption file directly. YouTube will sync it to your video and display captions automatically. This works even if you created the captions with WhisperX — YouTube doesn't care where they came from.

Caption files are plain text, so you can also open them in Notepad or any text editor and fix mistakes by hand. WhisperX is usually accurate, but it can mishear technical terms, names, or audio with background noise. Editing the file takes a few minutes for a long video, but it's faster than captioning from scratch.

Troubleshooting common problems

If you get an error saying "whisperx: command not found" or "whisperx is not recognized", Python or WhisperX didn't install correctly. Go back and run the pip install commands again, making sure you see "Successfully installed" at the end. On Windows, restart your computer after installing Python — sometimes the PATH doesn't update until you do.

If WhisperX starts but then crashes with an error about CUDA or GPU, your computer is trying to use a graphics card to speed things up, but something went wrong. Add --device cpu to your command to use your regular processor instead. It will be slower, but it will work: whisperx yourfile.mp4 --model base --language en --device cpu

If the captions are badly out of sync with the video, try a different model. The "tiny" model is faster but sometimes less accurate with timing. Switch to "base" or "small" and run it again.

If you're not sure whether your file is in the right folder, type dir (Windows) or ls (Mac/Linux) in your command prompt or terminal. It will list all the files in the current folder. Make sure your audio or video file is there before you run the WhisperX command.

Using a graphical interface instead of the command line

If typing commands feels uncomfortable, you don't have to. Faster-Whisper-Transcription is a free desktop process that does the same thing as WhisperX but with buttons and menus instead of text commands. You can read it from GitHub (search "Faster-Whisper-Transcription releases") and run it like any other program.

The trade-off is that a graphical process takes up more disk space and uses a bit more memory. For most people, it's worth it if the command line feels like a barrier. The captions it creates are identical to what WhisperX produces on the command line.

Another option is Subtitle Edit, which is a dedicated caption editor. It has WhisperX built in, so you can load a video file and generate captions without leaving the process. read it from subtitleedit.com.

Frequently Asked Questions

Can WhisperX caption video in languages other than English?

Yes. Add --language followed by the two-letter language code. For example, --language es for Spanish, --language fr for French, --language de for German, --language zh for Chinese. WhisperX supports over 90 languages. If you don't specify a language, it tries to guess, which usually works but isn't always reliable.

Do I need an internet connection to use WhisperX?

You need internet the first time you run it, because it downloads the language model (about 140 MB). After that, WhisperX works completely offline. Your audio files never leave your computer, and you don't need to be connected to the internet while it's captioning.

How accurate are the captions?

Accuracy depends on audio quality and the model you choose. Clear speech with no background noise usually gets 95% or better accuracy. Accented speech, technical terms, or noisy audio may have more mistakes. The larger models (small, medium) are more accurate than tiny or base, but they're slower and need more computer power. Most people find the captions good enough to use as-is, or quick to fix by hand.

Can I use WhisperX on a Mac with Apple Silicon (M1, M2, M3)?

Yes. Follow the same installation steps. WhisperX runs on Apple Silicon, though you may need to install some additional tools first. If you run into problems, search for "WhisperX Apple Silicon" — the community has documented workarounds. Alternatively, use a graphical process like Faster-Whisper-Transcription, which handles these details automatically.

What if I want to caption a live stream or a video I'm recording right now?

WhisperX works on files that are already saved to your computer, not on live streams. If you want captions while you're recording, you need different software — OBS Studio (free, open-source) has caption plugins, or you can use a live captioning service. After you finish recording and save the file, you can use WhisperX to create captions for the recording.