Paste a YouTube episode link and get the full text with a label on every speaker turn and a timestamp on every paragraph.
Trusted by 42,800+ creators
Sample coming soon
A YouTube podcast transcript is the full text of an episode that is on YouTube, with a label on each speaker and a timestamp on each paragraph. It is what you search, quote and paste into show notes.
We picked the three lines by hand. The tool gives you the transcript; you choose the lines.
Once you own the job, ask the built-in AI for a timestamped list of sections and paste it into the description as 0:00 Title. It reads the first 24,000 characters, about 25 to 35 minutes of speech.
Export the transcript as TXT or DOCX with speaker names and timestamps on or off. Share a read-only link to the text.
API and MCP are included on every plan. Your 10 free minutes work through the API too. Each call takes one YouTube URL and returns one transcript, so a weekly show is one call a week.
| Feature | YouTube auto captions | YoutubeToText |
|---|---|---|
| Speaker names | Dashes | Labels you can rename, free |
| Paragraphs | No | On every speaker change, at most every 30 seconds |
| Timestamps | Every few seconds | One per paragraph, toggle on or off |
| Filler words | Kept | Cleaned by default, verbatim on request |
| Summary | No | Paragraphs in the episode's language |
| Exports | None for videos you do not own | TXT and DOCX; SRT and VTT from a Subtitles (SRT or VTT) job |
| Episodes over 1 hour | Auto captions may not exist | Up to 12 hours |
Click a speaker chip and type the name. Rename once, applied everywhere.
Read the summary for the show notes draft. Back in Transcript, copy the lines you want with their timestamp.
Export .txt or .docx for the episode page. Run a Subtitles (SRT or VTT) job on the same link for SRT and VTT. Share the read-only text page with your editor.
A long episode is not one request. Videos over 10 minutes are split into chunks, each chunk is transcribed on its own, and the parts are merged back into one transcript. You get one transcript, not several. Speaker labels are tracked across the whole episode, so the host is the same speaker at minute 47 as at minute 2. The ceiling is 12 hours per video, which covers a full-day live show. The sample above is a full real episode, run once on our own account, and its length is stated next to the player. Play it from the middle and watch the active word follow the audio. If you need to check a line, click any word to jump the player to that second.
The summary is a set of paragraphs in the language of the episode. It has no bullet list and no timestamps, so it is a first draft of the show notes, not the finished post. Take the timestamped lines from the transcript yourself: open Transcript, find the moment, copy the paragraph with its timestamp. For chapters, YoutubeToText has no chapter output. Once you own the job, open Ask AI and ask for a list of the sections with a timestamp and a title, then reformat the answer as 0:00 Title in the YouTube description. Ask AI reads the first 24,000 characters of the transcript, about 25 to 35 minutes of speech, and answers in at most 700 tokens, so on a long episode ask it about the part you need.
YoutubeToText takes a YouTube link only. If your show is on Spotify, Apple Podcasts and YouTube, paste the YouTube upload. The video must be public or unlisted; a private video cannot be read. The transcript is the same whichever platform you publish the text on, so one job covers the episode page, the newsletter and the social posts. If the guest was recorded in a second language, run Translate on the job afterwards; the translation keeps the speaker labels and paragraphs. When you publish the full transcript on your site, credit the show, the episode, the host and link back to the episode, the same way you would quote any other source.
I still check for a transcript, and if there is none I move along. You can't search audio. You can't skim audio. You can't easily hyperlink to audio, copy-and-paste audio, or auto-summarize audio.
Right now I'm doing this manually by copy/pasting into ChatGPT but I want to automate this aspect.
10 min
Your first 10 minutes are free. They cover the start of a 90-minute episode; the rest is locked until you unlock it.
...
One payment for a 90-minute episode, 90 minutes. No subscription.
... / month
600 minutes every month, the smallest plan that covers a 90-minute episode.
...
600 minutes, one payment, credits never expire.
Updated September 2026
Ready when you are