Email →

Dec 2025 → Sep 2026 · prototype

Rokan Labs

A multi-month build where one sentence in a chat box comes out the other end as a published, captioned YouTube video. Nobody touches it.

What was wrong

Short-form content eats hours, and almost none of those hours are the creative part. It's the rendering, the captions, the hashtags, the uploading.

What I made

I built it at Halyard on n8n with Google Veo 3. You type an idea into a chat box. It refines the prompt, generates the video, writes the caption and hashtags off the finished video, and uploads it.

truefalseWebhookwebhookPrompt RefinementAI agentGenerate VideoHTTP requestWaitwaitCheck Video StatusHTTP requestIfifDownload DataHTTP requestTitle GeneratorAI agentSeparate Title and DescriptioncodeUpload a videoYouTubeHTTP RequestHTTP requestOpenAI Chat ModelOpenAI modelOpenAI Chat Model1OpenAI model

Rokan Labs · exported from n8n · 13 nodes · scroll sideways →

The actual workflow, exported from n8n. It refines the prompt, generates the video, then waits and checks until the render is done, titles it, and uploads it.

The hard part

It was a long build, multiple months, and the pieces had to hand off cleanly. One bad handoff and you get a published video with someone else's caption on it.

How it went

I pointed it at my own channel and it worked: I'd give it an idea and then just find a finished, captioned video on my YouTube. The plan was to make it a product.

How it works

  1. Prompt refinement: sharpens the idea you typed
  2. Video generation: turns the prompt into a video with Veo 3
  3. Caption: writes the caption and hashtags from the video
  4. Upload: posts it to YouTube

What came out of it

  • Four steps, end to end, with no human in between
  • Built with n8n and Google Veo 3

Status · prototype

Built during my AI automation internship at Halyard Consulting. I've since left to run my own automation business. rokanlabs.com is still up, but the pipeline behind it is off, so the Short doesn't get generated anymore.