Developer built a video transcriber in 30 minutes using vibe-coding in Codex

A developer assembled a video transcriber through vibe-coding in Codex in 30 minutes

The author of a post on Habr described how, using vibe-coding in Codex, they assembled a working tool for transcribing video recordings of meetings — the first version was written in approximately 30 minutes.

Why the tool was needed

After work meetings, the author has video recordings — in their case, recordings from the Yandex Telemost service, and simply rewatching them is inconvenient. If something needs to be recalled, you have to open the video, rewind it, search for the right moment again, and then listen to that fragment once more. Another frequent scenario the author encounters regularly is passing context about a task to a programmer: you can send the meeting recording and write "please watch," but then the person would have to watch the entire video themselves and figure out where in the conversation the discussion is, where the decision was made, and where the task was formulated.

The author wanted a more convenient option: to send not only the recording itself, but also a text summary of the meeting along with it.

• Source of recordings — Yandex Telemost service
• Problem — finding the right moment requires rewinding and listening through the entire video
• Use scenario — passing meeting context to a programmer regarding a task
• First working version of the tool assembled in approximately 30 minutes
• Tool assembled through Codex using the vibe-coding method

What the tool does

According to the author, the idea was simple: upload a meeting video and receive a transcript, protocol, and DOCX file that can be sent instead of the recording itself. Three text artifacts — a transcript of the conversation, a brief meeting protocol, and a ready DOCX document — replace the need to rewatch the original video in full: the recipient only needs to read the text to understand what was discussed and what needs to be done, without opening the video file itself.

"The first working version was assembled through

Codex in approximately 30 minutes" — from the author's post on Habr.

"The idea was simple: upload a meeting video and get a transcript, protocol, and DOCX that can be sent further" — the author of the material,

Habr.

How the prototype was assembled quickly

The first working version took approximately 30 minutes to assemble — exactly how long, according to the author, it took to get a working tool through Codex instead of writing code from scratch manually. Vibe-coding in this case means that the author formulated the task and edits through dialogue with an AI tool, rather than writing each line of code independently, which allowed them to fit the development into a half-hour timeframe from idea to working prototype. According to the author's description, the path from task formulation ("upload video — get transcript, protocol, and DOCX") to the first working version took exactly this half-hour period, without preliminary infrastructure setup or a separate design phase.

Who would find such a tool useful

The author describes the application primarily for passing context to a programmer regarding a specific task — this is the case when simply sending a video is insufficient, because the recipient would still have to spend time watching. With a ready transcript, protocol, and DOCX file, the tool does this work for the recipient: the colleague reads the summary of the meeting instead of watching the entire recording, and the original Yandex Telemost video remains 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 ready personal tool: instead of searching for a third-party transcription service, the author assembled a narrowly specialized solution for their own workflow in approximately 30 minutes — Yandex Telemost recording on input, transcript, protocol, and DOCX on output.