Read a long video as a few paragraphs, then open the transcript under it and check any line against the second it was said.
Trusted by 42,800+ creators
Sample video: Dating Dilemmas: Our Wild Experiences & Major Red Flags | EPISODE 4, 68 minutes
In this episode of their podcast, hosts Martina and Ajib celebrate their one-month anniversary and delve into the complexities of the dating scene in Nairobi. They deeply analyze the challenges of dating young, especially when the relationship is in the public eye. Both hosts reflect on their past highly public relationships, noting the immense pressure that comes from social media and invested fans. They share how public breakups are complicated by factors like co-parenting, managing joint businesses, and dealing with public scrutiny. The hosts emphasize the importance of giving grace to their younger selves, acknowledging that early relationships serve as essential life lessons rather than wasted time.
0:00Martina:Hey, guys. This is Martina.
0:02Ajib:And this is A jib on "Two Peas in a Pod"
0:04Martina:Cast, and this is episode four. [upbeat music] Hey. Let's go.
0:19Ajib:Beautiful intro.
Switch to Transcript and the same episode is there in full, with a timestamp on every paragraph. That is the part the summary is built from, and the part you check it against.
“Creates text summarise of long YouTube podcasts that I can feed into AI and get a quick summary of.”
A YouTube video summarizer turns the spoken words of a video into a short piece of prose you can read in a minute. This one writes the summary from a full transcript and keeps that transcript, so every claim in the summary can be traced back to the moment in the video it came from.
In this episode of their podcast, hosts Martina and Ajib celebrate their one-month anniversary and delve into the complexities of the dating scene in Nairobi. They deeply analyze the challenges of dating young, especially when the relationship is in the public eye. Both hosts reflect on their past highly public relationships, noting the immense pressure that comes from social media and invested fans. They share how public breakups are complicated by factors like co-parenting, managing joint businesses, and dealing with public scrutiny. The hosts emphasize the importance of giving grace to their younger selves, acknowledging that early relationships serve as essential life lessons rather than wasted time.
0:00Martina:Hey, guys. This is Martina.
0:02Ajib:And this is A jib on "Two Peas in a Pod"
0:04Martina:Cast, and this is episode four. [upbeat music] Hey. Let's go.
Real output of one sample job, unedited: the first paragraph of the summary, then the first three paragraphs of the transcript with the times they were said. Nothing here was written by hand.
A summary you cannot verify is a guess you have to trust. The transcript stays in the next tab, so you read a claim, click into the transcript and hear it said. Nothing is thrown away once the summary is written.
The summary comes back in the language that is spoken in the video, not in English by default. A Polish interview is summarized in Polish, a Japanese briefing in Japanese.
A video can run to 12 hours. Long ones are transcribed in chunks and come back as one transcript and one summary, so a three-hour meeting does not have to be cut up by hand first.
| Feature | Copy into a chat tool | YoutubeToText |
|---|---|---|
| Getting the text | Copy YouTube's panel, when the video has one | The audio is transcribed, captions not needed |
| A long video | You split it and paste it in parts | One job, up to 12 hours |
| Checking a line | The summary is all you kept | The transcript stays, timestamped |
| Speakers | Lost in the copy | Labelled, and you can rename them |
| Cost of the summary | A chat subscription | No minutes on top of the transcript |
The engine listens to the audio of the video and writes it out, so it works on videos that have no captions and on ones whose auto captions are poor.
Click Generate summary. One request goes out and a few paragraphs come back, under a single Summary heading, in the language spoken in the video.
Switch to Transcript and click any timestamp to jump the video to that paragraph. This is the step the summary alone cannot give you.
Copy the summary straight into your notes, or export the transcript as Word or plain text, with speakers and timestamps on or off.
The summary is a few paragraphs under one Summary heading. It has no bullet points, no section headings and no timestamps, because the timestamps are stripped out of the transcript before the summary is written and the instruction is to write paragraphs rather than lists. That is the right shape for reading and the wrong shape for navigating. To navigate, use the transcript: every paragraph carries a time, and Find shows you each place a term was said with its timestamp beside it. If you want chapters or an outline with times, ask the built-in AI on a job you own. It reads the first 24,000 characters of the transcript, which is roughly 25 to 35 minutes of speech, so on a long video it sees the opening and not the end.
The summary is written in the language of the text it is given, and the order you work in decides what that text is. Summarize first and you get the summary in the language spoken in the video. Set a target language first and the transcript is translated, so the summary is written from the translation and comes back in the target language instead. Neither one is wrong, but they are not the same, and switching the target language afterwards does not rewrite a summary you already have. If you want both, generate the summary before you translate, then translate.
A new account starts with 10 free minutes. Paste a two-hour video and you do not get an error: you get a preview covering as many minutes as you have left, with the transcript, the timestamps and the player working on that part. Generate a summary there and it summarizes the preview, because the preview is all the text there is. It will read as a fair summary of the opening and say nothing about the rest, which is easy to mistake for a summary of the whole video. Unlock the video, or take a plan, and the job runs from the start. Generate the summary again afterwards to get one that covers all of it.
It appears to have captured the gist of the meeting very well, but the fact that the transcript isn't close to an exact match to what was said - and the timestamps are incorrect - means it's very hard to trust the output.
I'm trying to do something technically similar by using LLMs to summarize the meeting transcript from youtube. Right now I'm doing this manually by copy/pasting into ChatGPT but I want to automate this aspect.
4 ways to pay for the same job. Start free, pay for the one episode, or cover a run of them.
10 min
Trying it on your own episode
Your first 10 minutes are free. They cover the start of a 1-hour video; the rest is locked until you unlock it.
...
One episode, nothing to cancel
One payment for a 1-hour video, 60 minutes. No subscription.
... / month
A show that publishes every week
600 minutes every month, the smallest plan that covers a 1-hour video.
...
A back catalogue in one go
600 minutes, one payment, credits never expire.
Not satisfied? We refund your purchase within 24 hours, no questions asked.
Updated August 2026
Ready when you are