How to turn video, audio and documents into text

Use a YouTube or other supported video link, an audio link, or a file you already have: a recording, a PDF, a scan or a photo of a page. Read the transcript alongside playback or the original page, then copy the words you need or download the text.

1. Copy the link

   Copy the address of a video or audio on YouTube, Vimeo, Rumble, Odysee or one of the many other sites TranscriptFox reads.

2. Paste it

   Paste the link into the box on the home page. TranscriptFox starts right away.

3. Read, copy or download

   The first 10 minutes appear as plain paragraphs on Free. Copy the available text with one click, download .txt, or unlock .srt, .vtt, .pdf and Word .docx exports.

Captions and audio transcription

TranscriptFox uses the captions that come with the video or audio. It prefers captions in the spoken language, and captions written by people over automatic ones. When a site does not say which language is spoken, English captions come first. If there are none, it listens to the audio and writes the transcript with speech recognition. That takes a little longer, depending on the recording length and current demand.

Transcripts keep the spoken words. Sound labels in square brackets, such as [Music] or [Applause], and music note symbols are left out. Where the captions mark a new speaker, a new paragraph starts, but speakers are not named, and transcripts made from the audio do not mark speaker changes.

Check the transcript against the recording

Under the title, TranscriptFox says where the text came from: Published captions is a caption or transcript track the site publishes with the recording (sites do not always say whether a person or a machine wrote it), Automatic captions are made by the site's own speech recognition, Transcribed from the audio means our speech recognition, and Transcribed from the audio (premium) marks Plus and Pro speech transcripts within your plan's monthly audio allowance. Speech recognition across Free, $1 Pass, Plus and Pro uses our high-accuracy speech engine by default. On Plus and Pro, when a recording's published captions are complete, they are shown as Creator’s captions, and Use premium speech transcribes the audio with speech recognition instead. When our checks detect an unreliable part, the transcript shows a marked gap with its start and end time. These checks cannot detect every error or missing word. Transcripts are not word-for-word records of every sound: filler words such as um and uh, false starts and background sounds may be left out, and sites often edit their published captions. Ads in a podcast's audio are transcribed with the rest. The downloaded text file starts with the title, the link and this label. The first transcript of a link is kept and shown to anyone who pastes that link later, so it reflects the recording as it was then, even if the source was edited or removed since. Transcripts are not certified: check quotes, names and times against the recording before you publish or rely on them.

Music and sung words

Songs and singing are transcribed too. With captions, sung words appear as the captions show them. With speech recognition, sung words are included when they can be made out reliably, and a part our checks reject is marked as a gap. Music without words adds no text from our speech recognition, though a site's automatic captions sometimes put a stray word such as “foreign” there. A recording with only music or silence shows a message instead of a transcript. Lyrics appear as paragraphs, not line by line.

Languages and transcription accuracy

TranscriptFox is designed to transcribe the original language, not translate it. Caption tracks identified as translations are skipped, and if a recording has only translated captions, TranscriptFox transcribes the audio instead. Some sites do not say which language is spoken or which caption track is the original, so a translated track can still be used now and then. Known automatic translation tracks from YouTube and TikTok are skipped on new requests. Speech recognition can miss words or translate a passage by mistake, especially when people switch languages, speak over each other, speak quietly or are hard to hear over noise. A missing passage may have no gap marker. Accuracy depends on the captions and the audio quality; turn on Timestamps to check the recording. Older cached transcripts can retain errors from earlier processing.

Free previews and full transcripts

The free plan includes 1 transcript a day without an account, or 10 a day when signed in, shared across the website, API and MCP. Daily allowances reset at 00:00 UTC. A $1 Pass unlocks 1 full transcript up to 2 hours with .txt, .srt and .vtt downloads. Plus ($5/month) gives 100 full-length transcripts a day and 30-day saved history. Pro ($20/month) gives 500 full transcripts a day, permanent saved library with full-text search, .pdf and .docx exports and batch transcription. Audio transcription is limited to 4 hours and 600 MB per recording. See plans.

Upload a file

If a website cannot share its media with TranscriptFox, you can upload a video or audio file you already have. Choose “Upload audio, video, PDF or image” on the home page, or drag the file onto the page. Recordings can be up to 600 MB and four hours long. On the free plan, uploads also show only the first 10 minutes and count toward your 10 transcripts a day when signed in. Service capacity and daily speech safeguards still apply.

PDFs, scans, photos and handwriting

Upload a PDF, or a JPEG, PNG, WebP or TIFF image, up to 600 MB. Text stored in a PDF is read directly, page by page and in reading order, including two-column pages and simple tables. Scanned pages and photos are read with OCR, including sideways and upside-down pages. Handwriting is read twice by two different handwriting models on TranscriptFox’s own servers. Words the readings disagree on are highlighted, and words that cannot be read show as [illegible]. Text is transcribed as written: spelling, numbers, names and punctuation are not corrected, and nothing is summarized. Handwriting recognition still makes mistakes, especially on messy or faint writing, so check highlighted words against the original.

Each document counts as one transcript. Free shows the first 3 pages. A $1 Pass reads up to 100 pages, Plus up to 200 and Pro up to 500 pages per document. Handwriting is read on up to 50 pages per document, and reading stops after 15 minutes per document; split a very long file to read the rest. Password-protected PDFs and PDFs whose owner blocked copying text are refused.

Use the Find box above a document to search it, and Page numbers to show or hide page headings in the text, in Copy and in downloads. Show original opens the file from your own device beside the text, at the page you click; it is not uploaded again. Edit text shows each passage as an editable box: fix any word, then Copy, downloads and Save use your corrections. The file is deleted after processing, and no copy of it is kept.

Listen and check timestamps

For a YouTube link, the video itself plays in YouTube’s own player below the title. For other links, use Play below the title to listen to the audio, pause, seek, mute or change the volume. Audio from a link may take a moment to prepare. Playback is available for accessible recordings up to four hours long. Audio from a link is kept for up to one hour; press Play to load it again. Audio of an uploaded file is kept for 24 hours if you are not signed in, 30 days on a Free account and while you subscribe to Plus or Pro. Delete audio below the player removes it sooner. After that, upload the file again to listen.

Copy, download and print

To print a transcript, use your browser’s Print command. The printout has the title, recording details and transcript, without the player and buttons. Use Copy to paste the text into another app.

Use your browser’s Find command to locate a word or phrase in the transcript shown on the page; documents also have their own Find box. Copy or download the available text in .txt, .srt, .vtt, .pdf or .docx format depending on your plan. Free previews include only the first 10 minutes of a recording or the first 3 pages of a document.