Save a Recipe From a YouTube Video, and Know When It Will Not Work
Rose · 19 September 2026
A twelve minute video. Someone walks you through a ragu properly, the way a written recipe never quite does, with the pan on screen while they talk. Somewhere in there are the amounts. There is no recipe in the description, just a chapter list, three affiliate links and a request to subscribe.
You are going to cook it. You are not going to scrub back and forth through twelve minutes with a notepad to find out how much wine went in at minute four. So the useful question is not whether a video can be turned into a recipe, it is which videos can, and how you tell before you commit an evening to one.
Where the recipe actually is in a cooking video
On a food blog the recipe is written down and the video is decoration. On YouTube it is usually the other way round. The amounts are spoken over the footage, sometimes flashed on screen for a second, and the description underneath is written for the algorithm rather than for you: chapters, links, a short pitch. The recipe exists, it is just stored as speech.
That matters because it decides what any tool has to work with. Anything that reads a page will find the description, which is not the recipe. Anything that reads the recipe has to read the talking.
What YouTube already gives you for free
Before any app, YouTube itself will show you the words. Open the description and there is a transcript, the same text the captions use, timestamped line by line. You can read it, scroll it and copy from it, which already beats rewinding.
Those captions are often not written by the creator. YouTube generates them with speech recognition, and its own help page is blunt about the consequence: the quality varies, and captions can misrepresent what was said because of mispronunciations, accents, dialects or background noise. A hot pan is background noise. So the transcript is a real head start and not a finished recipe, and the amounts are exactly the part most worth double checking against the footage.
The transcript also arrives as one long ribbon of talking, with the chat, the sponsor read and the story about the creator’s grandmother mixed in with the cooking. Turning that into an ingredient list and a set of steps is still a job. It is just a much smaller job than transcribing by ear.
The five times there is nothing to read
Sometimes the transcript is simply not there, and it is worth knowing why, because the reasons are documented rather than mysterious. YouTube lists them:
- The captions are not ready yet, because the audio is complex and still processing.
- Automatic captions do not support the language spoken in the video.
- The sound quality is poor, or the speech is not recognised.
- There is a long stretch of silence at the start.
- Several people talk over each other, or several languages run at once.
Read that list next to a cooking channel and the overlap is obvious. Two people cooking together and talking across each other is a format, not an accident. A long silent beauty shot of the ingredients laid out before anyone says a word is a format too. A video can be perfectly watchable, clearly audible to you, and still carry no caption track at all.
This is the part that tends to get skipped when apps describe themselves as working with YouTube. The link always works. The reading does not always work, and the difference between those two sentences is where an evening goes.
What Yumzi does with a YouTube link
Send it a video and it reads the caption track as the source, because that is where the recipe is. A normal watch link, a youtu.be link, a Shorts link and an embed link all resolve to the same video, so it does not matter which one you happened to copy. What comes back is an ingredient list and steps you can check against the footage before you shop, with the creator’s channel credited on the card and their thumbnail on it.
Under the ingredients there is a line that says the recipe was read from the video, so the ingredients can be incomplete. That line is not hedging for its own sake. A video is a lossy place to keep a recipe, the reading is only as good as the caption track underneath it, and you deserve to know which kind of source you are looking at while you are still at the shopping stage.
And when the caption track is empty, you get told the video could not be read. Nothing is generated to fill the gap. This is the part worth being plain about, because it is the one place where the tempting behaviour and the correct behaviour point in opposite directions: an app that produces a confident ingredient list from a video nobody could read has not solved your problem, it has moved the failure from the screen to the kitchen, where it costs you a shop and a dinner. A recipe that says it could not read something is more useful than one that guesses.
What to do with a video that will not read
If the transcript is empty, the recipe is still in there, it is just going to cost you the twelve minutes. A few things make that cheaper.
Check the description properly first, all of it, not the two lines shown before the fold. A fair number of channels do write the recipe out and bury it under the chapter list and the affiliate block. Check the pinned comment for the same reason. Check whether the channel has a site, since creators who put this much work into a video often keep a written index somewhere, organised by dish.
If none of that exists, watch it once at normal speed without stopping, then once more with the chapter markers, which usually land on the actual steps and let you skip the story. Write down the amounts on the second pass only. And if the video is one of two people talking over each other, accept early that the transcript was never going to exist and that you are doing this by hand.
The short version
A YouTube recipe lives in the talking. If the video has a caption track, that talking can be lifted out and turned into something you can shop from, by you with the transcript open or by an app that does the reading. If it does not, no tool can read it, and the only honest thing for any of them to do is say so.
Facts about YouTube captions in this piece were checked on 19 September 2026. Next check: 19 December 2026.
FAQ
Does this work on YouTube Shorts as well as long videos?
Yes. A Shorts link, a normal watch link, a youtu.be short link and an embed link all point at the same video, and all four resolve to the same video id. What decides whether the recipe comes across is not the length or the link format, it is whether the video has a caption track to read.
What happens if the video has no captions at all?
You get told the video could not be read. That is the whole answer, and it is deliberate. A video with no caption track gives back no text, and the honest response to no text is to say so rather than to produce a plausible looking ingredient list that nobody actually said. An invented amount is worse than no recipe, because you only find out at the pan.
Why does a video I can clearly hear still come back empty?
Because the reading works from YouTube's caption track, not from the audio in your ears. YouTube generates those captions itself, and its own documentation lists the cases where it does not: the language is not supported, the sound quality is poor, several people talk over each other, there are multiple languages at once, there is a long silence at the start, or the captions are still processing. In all of those the video plays perfectly and the caption track is empty.
Yumzi is on the App Store.
Download it and cook your saved reels this week.