The most common failure of every YouTube transcript tool is the same: the video has no caption track, and the tool gives up. Our transcriber closes that gap. It downloads the video's audio and runs Whisper-class speech recognition on it, producing a brand-new timestamped transcript you can read, copy, or export.
Captions first, transcription as fallback
The tool is smart about what it does first. If the video has an official caption track, it extracts that — it is the exact source text and costs the least. Only when captions are missing (or you explicitly force it) does it fall back to transcribing the audio. Either way you get the same output shape: segments with start time, duration, and text.
How the async job works
- Paste the video URL. If captions exist, you get the result immediately.
- If not, the tool starts an ASR job and returns a job ID in a second or two.
- The job polls automatically in the browser (or via
/api/asr/jobfrom your code) until it is done, then the transcript appears with all the usual export buttons.
What you can export
- TXT — one line per segment with
[mm:ss]timestamps. - SRT / VTT — real subtitle files with frame-accurate timings, ready for editors and players.
- JSON — machine-readable segments (start, duration, text) for pipelines and LLM context.
Typical uses
- Podcasts and talks uploaded without captions
- Lectures and courses — searchable study notes
- Interviews and webinars — turn hours of audio into citable text
- Accessibility — add subtitles to videos that lack them
- LLM pipelines — transcripts instead of raw audio for summarization and RAG
Related
The specific case — a video with no caption track at all — is covered on the transcript for a video without subtitles page. If your content is full of jargon that speech models mangle, see transcription with a custom word list. To feed transcripts into an LLM or RAG pipeline, see transcripts for LLMs. For the whole product — batch, translation, summaries, and the API — start with the free tool.
Frequently asked questions
Does it work on videos that have no captions at all?
Yes — that is the whole point. The audio track is downloaded and transcribed with a Whisper-class speech recognition model, producing a new timestamped transcript even when YouTube has no caption data for the video.
How long does transcription take?
A typical 5–10 minute video transcribes in a few minutes, depending on audio length. The job runs asynchronously: you get a job ID immediately and can poll its status from the browser or the API.
How accurate is the transcription?
For clear speech in common languages, Whisper-class models are very accurate (typically high-90s word accuracy). Heavy accents, music, or multiple overlapping speakers reduce accuracy.
Can I transcribe a video and then translate it?
Yes. Once the transcript exists, you can request it in any supported language and download the translation in the same formats.
Which formats can I export?
TXT (timestamped lines), SRT and VTT subtitles, and raw JSON with per-segment start/duration.
Can I bias the transcription toward my own terms (jargon, product names)?
Yes — the local-file upload accepts a custom word list. You type the terms that keep getting mangled (one per line) and the speech model is biased toward your exact spelling via Whisper's initial prompt. Save the list with your account and it applies to every upload automatically. Details on the custom word list page.