How to transcribe a video to text for free

Transcribing a video means writing down what's said in it. The same words can come out in a few forms, depending on what you need them for:

They all start the same way: speech recognition, or someone typing. Once you have the words with their timings, turning them into any of the three is quick.

In your browser. The free video to text converter takes a video or audio file, transcribes it on your device and gives you the text in paragraphs with timestamps, to copy or download as TXT, SRT or VTT. The first time, your browser downloads the speech model and keeps it, so the next video starts in seconds. It's quickest on a computer, and faster still with a recent graphics chip.

On YouTube. If the video is yours and already on YouTube, it probably has automatic captions. In YouTube Studio, open Subtitles, pick the video, choose Edit next to the language, then Options and Download subtitles. You get a captions file with timings; to correct the automatic ones first, choose Duplicate and edit. YouTube says some videos don't get automatic captions: very long ones, ones with poor sound, a long silence at the start, or people talking over each other.

Uploading to a transcription service works too, and many have a free allowance. Your file goes to their servers, so check what they keep and for how long before you upload anything private.

1. Open the video to text converter and choose your video, or an audio file. MP4, MOV and WebM work.
2. Pick the language, or leave it on automatic. Naming it helps with short clips and strong accents.
3. Transcribe. The first run downloads the speech model; after that it loads from your browser in seconds. Long videos take a few minutes.
4. Read it through and fix it: names, numbers and jargon first (below).
5. Copy it or download it, as plain text, or as SRT or VTT with the timings for captions.

Speech recognition is good at clear speech and stumbles where people would too. YouTube lists the usual causes for its own automatic captions: mispronunciations, accents, dialects and background noise. In practice:

• The sound matters most. A microphone close to the speaker beats a camera across the room. Music under the voice and a noisy room cost the most words.
• One voice at a time. People talking over each other is where transcripts fall apart; the words may come out, but not in order.
• Names, brands and numbers are the usual mistakes: the model writes what a word sounds like, so an unusual name comes out as a common word. Search for each guest's and product's name and fix every one.
• Say which language. Detection guesses from the first seconds, so a video that opens with music or another language can be misread from the start.

A timestamp at the start of each paragraph, like [12:40], makes a long transcript easy to find your way around, and lets anyone jump to the moment in the video. Timestamps are also how YouTube chapters work: list them in the description, starting at 00:00, with at least three in order and each chapter 10 seconds or longer.

Captions need timings for every line, not every paragraph. When the transcription times each word, the words can be grouped into short lines with exact start and end times: an SRT or VTT file. The free auto caption generator does that and lets you fix each line, and how to add subtitles to a video covers putting them on YouTube, in an editor or into the video itself.

• A blog post or article. Edit the spoken words into readable paragraphs: cut the filler, add headings, and the video's ideas get found in search.
• Show notes and chapters. Pull out the topics with their timestamps for a podcast's show notes or the video's description.
• Quotes. The best line in an episode makes a post on its own, as text or as a quote card.
• The clip-worthy moments. Read a long video's transcript and the moments that would work as clips stand out: the strong opening line, the story with a payoff, the answer that lands. That's exactly how clips are found; see how to repurpose video content and how to clip a podcast.