Auto subtitle generator that runs on your computer

Subtitles without typing them

An auto subtitle generator listens to the speech in a video and writes it out as subtitles: each line of text, with the moment it appears and the moment it disappears. You don't type anything. You check the result and fix what the AI got wrong.

Wolfcaps does this inside your browser. The speech recognition model runs on your computer's own GPU, so your video is never uploaded anywhere. It's free, there's no account, and nothing is stamped on your video.

How to make subtitles automatically

1. 1Open Video → subtitles and drop your video, or choose it from your computer. If it's on your phone, click Send from your phone and scan the QR code.
2. 2If the video already has subtitles inside it, Wolfcaps finds them and uses them instead of transcribing. You can still transcribe the speech if you prefer.
3. 3In Transcribe this video?, pick the spoken language (English, Hebrew or Russian) and a speech model: Balanced or Best. The first time, the model downloads once and stays in your browser.
4. 4Wolfcaps measures your device's speed, shows how long the job will take, and starts. Subtitles appear in the list as they're made.
5. 5Click a line to hear it and fix its text. For timing, splitting and merging, open the Full editor.
6. 6Click Export… and save the subtitles as a file, or the video with the subtitles inside.

What you get

• Subtitle files.

  SRT (works in every player), VTT (web players, YouTube, course platforms), ASS (styled subtitles for VLC, mpv and MKV) or TXT (just the words, in paragraphs).

• The video with subtitles inside.

  An MP4 or MKV with subtitle tracks the viewer can turn on. Your video and sound are copied, not re-encoded, so it's fast and there's no quality loss.

• Translations.

  You can translate the subtitles into 20 languages and put several languages in one video.

Fixing what the AI gets wrong

Automatic subtitles are a first draft. Names, rare words and mumbled lines are where mistakes usually show up. Wolfcaps gives you the tools to fix them quickly:

• Click any line to hear it again and correct the text.
• Adjust timing on a waveform, split a long line, merge two short ones, insert or delete lines.
• Use find & replace to fix a name everywhere at once.
• Run the quality check. It flags lines that are too fast to read, too long or overlapping the next one, and offers fixes.
• Undo and redo anything.

When you leave, you can keep the project in your browser. Kept projects are listed under Recent projects on the home page, so you can come back to them later.

Speed and long videos

On a recent desktop GPU, Wolfcaps runs about 10 times faster than real time: a 10-minute English video takes about a minute. Slower GPUs take longer. Before starting, Wolfcaps runs a short speed test on your device and shows an estimate, so you know what to expect.

Balanced is the faster model, with near-best accuracy. Best is slower and more accurate. For Hebrew, Best is a bigger model (about 2.1 GB) that does much better on hard audio. Hebrew uses ivrit.ai's Hebrew Whisper models, converted by us to run in the browser.

Multi-hour files work. If you stop partway, keep the project: it waits in Recent projects and continues from where it stopped.

What it can't do yet

• Speech recognition works for English, Hebrew and Russian only. Videos in other spoken languages can't be transcribed yet.
• You need a computer with Chrome, Edge, Brave or another Chromium browser, or Safari 26 or newer on a Mac, and a GPU the browser can use (WebGPU). Firefox, iPhone and iPad aren't supported yet.
• The speech model is a one-time download of about 1.6 GB (2.1 GB for Hebrew Best), kept in your browser.