Developer builds a video meeting transcriber in 30 minutes with vibe coding in Codex
A Habr author built a working transcriber for recorded meetings using vibe coding in Codex. The first version took about 30 minutes. The idea is simple: you upload a recording (for example, from Yandex Telemost) and get a transcript, meeting minutes and a DOCX file you can send to a colleague instead of making them watch the whole video. The author says the tool was built through a dialogue with the AI, without writing every line of code by hand.
AI-processed from Habr AI; edited by Hamidun News
A Habr author described how he used vibe coding in Codex to build a working tool for transcribing recorded meetings. He wrote the first version in about 30 minutes.
Why the tool was needed
After work meetings, the author is left with video recordings. In his case they come from Yandex Telemost, and simply rewatching them is inconvenient. To recall something, he has to open the video, scrub through it and hunt for the right spot, then listen to that fragment again.
Another frequent scenario he runs into: passing on the context of a task to a programmer. You can send the meeting recording and write "please take a look," but then the person has to watch the entire video and work out for themselves where the discussion is, where the decision was made and where the task was formulated. The author wanted a more convenient option: forwarding not only the recording itself but also a text summary of the meeting along with it.
- Source of recordings: the Yandex Telemost service
- Problem: finding the right moment requires scrubbing through and re-listening to the whole video
- Use case: passing meeting context to a programmer for a task
- The first working version of the tool was built in about 30 minutes
- The tool was built in Codex using vibe coding
What the tool does
The idea, according to the author, was simple: upload a meeting video and get a transcript, minutes and a DOCX file that can be sent on instead of the recording itself. Three text artifacts (a transcript of the conversation, brief meeting minutes and a ready-made DOCX document) remove the need to rewatch the original video in full. The recipient only has to read the text to understand what was discussed and what needs to be done, without opening the video file.
"I managed to build the first working version in
Codex in about 30 minutes" — from the author's post on Habr.
"The idea was simple: upload a meeting video and get a transcript, minutes and a DOCX that can be sent on" — the author of the piece,
Habr.
How quickly the prototype came together
The first working version took about 30 minutes. That is how long, by the author's account, it took to get a working tool through Codex instead of writing the code by hand from scratch. Here, vibe coding means the author phrased the task and the corrections as a dialogue with the AI tool rather than writing every line of code himself, which let him fit into a half-hour window from idea to working prototype. According to the author, the path from stating the task ("upload a video, get a transcript, minutes and a DOCX") to the first working version took exactly that half hour, with no prior infrastructure setup and no separate design stage.
Who would find such a tool useful
The author describes the use primarily as a way to pass context to a programmer for a specific task. This is the case where simply sending a video is not enough, because the recipient still has to spend time watching it. With a ready transcript, minutes and DOCX file, the tool does that work for the recipient: the other person reads the outcome of the meeting instead of watching the whole recording, and the original Yandex Telemost video stays available as a source if details are needed.
What this means
The example shows how vibe coding shortens the path from an everyday work problem to a finished personal tool. Instead of looking for a ready-made third-party transcription service, the author built a narrowly specialized solution for his own workflow in about 30 minutes: a Yandex Telemost recording goes in, and a transcript, minutes and a DOCX come out.
- Nov 1, 2026OpenAI Rewrites GPT Behavior Rules: Five Values and a Chain of Command
- Nov 1, 2026Local AI models agreed to pay scammers, while simple rules blocked all 42 schemes
- Oct 31, 2026Author puts his script before an AI jury: the verdict cost 850 thousand tokens
- Oct 31, 2026Anthropic: Fold Subagents Back Into a Single Loop — One Customer's Tasks Ran Twice as Fast
Key Questions Answered
Why the tool was needed?
After work meetings, the author is left with video recordings. In his case they come from Yandex Telemost, and simply rewatching them is inconvenient. To recall something, he has to open the video, scrub through it and hunt for the right spot, then listen to that fragment again. Another frequent scenario he runs into: passing on the context of a task to a programmer. You can send the meeting recording and write "please take a look," but then the person has to watch the entire video and work out for themselves where the discussion is, where the decision was made and where th
What the tool does?
The idea, according to the author, was simple: upload a meeting video and get a transcript, minutes and a DOCX file that can be sent on instead of the recording itself. Three text artifacts (a transcript of the conversation, brief meeting minutes and a ready-made DOCX document) remove the need to rewatch the original video in full. The recipient only has to read the text to understand what was discussed and what needs to be done, without opening the video file. "I managed to build the first working version in Codex in about 30 minutes" — from the author's post on Habr. "The
How quickly the prototype came together?
The first working version took about 30 minutes. That is how long, by the author's account, it took to get a working tool through Codex instead of writing the code by hand from scratch. Here, vibe coding means the author phrased the task and the corrections as a dialogue with the AI tool rather than writing every line of code himself, which let him fit into a half-hour window from idea to working prototype. According to the author, the path from stating the task ("upload a video, get a transcript, minutes and a DOCX") to the first working version took exactly that half hour
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.