Transcribe Speech From Any Video Into Text on Your Own Device

Turn the speech in a video into text

This video to text converter listens to the audio in your video and writes down what people say, on your own device. You don't upload anything or make an account. You drop in a file and get readable text back, split into paragraphs so you can paste it into a document straight away.

The transcription comes from a real speech recognition model: NVIDIA's multilingual FastConformer, run through sherpa-onnx. Your browser downloads it once, caches it, and runs it locally for every video after that. It understands ten languages without being told which one it's hearing, which is the main difference from my MP4 to transcript tool. That one uses a separate model per language and exports timed subtitles.

How to convert video to text

Which transcription tool should you use?

I run two transcription tools on this site because they're built on different models and suit different jobs:

Works with video and audio files

Any video with a soundtrack works: an MP4 from a phone or a download, a MOV from an iPhone, an MKV, or a WebM screen recording. The tool only reads the audio track, so the resolution and picture quality don't matter. A shaky phone clip and a clean studio recording are read the same way.

You can also drop in an audio file directly, like an MP3 voice memo or a podcast export, with no video around it. If a file has no audio track at all, the tool tells you instead of returning an empty page.

What people use it for

• Meeting recordings. A Zoom, Teams or Google Meet call saved as video becomes text you can search, skim or paste into notes, instead of an hour of footage to scrub through.
• Lectures and talks. A recorded class or conference session becomes something you can search and study from.
• Interviews. A recorded conversation becomes a written draft for an article, a research write-up or show notes.
• Clips you want to quote. A webinar, a talk or a video message where you need the exact words that were said.

Timestamps, copying and downloading

By default you get clean paragraphs with no timestamps, which is what you want for pasting into a document. Turn on "Show timestamps" and the same transcript splits into timed lines, so you can jump to the moment something was said.

Either version copies to your clipboard or downloads as a plain .txt file. The download is text, not a subtitle file. If you need an SRT or VTT file to add captions to the video, use the MP4 to transcript tool, which exports both. I wrote about the difference between a transcript and subtitles if you're not sure which one you need.

Getting a better transcript

Accuracy depends mostly on the audio:

• Clear, close speech works best. A call recording or a video with a clip-on mic transcribes far better than a phone recording from across a room.
• One speaker at a time helps. People talking over each other is the hardest case for any transcriber, human or machine.
• Background music and noise cost accuracy. A quiet meeting transcribes better than the same conversation with music underneath.
• Long videos take longer. There's no length cutoff, but a two-hour recording takes much longer than a two-minute clip. Keep the tab open and in the foreground while it works.

No automatic transcript is word-perfect on difficult audio, and this one isn't either. Treat the result as a good first draft to skim and correct, especially for anything where exact wording matters, like legal or medical recordings.

If you want to know what the model does with your audio, step by step, I explained it in how browser speech-to-text works.

Frequently asked questions