I built a YouTube for generative videos in 50 prompts

Name: I built a YouTube for generative videos in 50 prompts
Availability: InStock
Author: proy24

by proy24·Jul 11, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●MidShip It

YouTube for AI videos when YouTube already hosts generative content.

Strengths

•MIT open-source license allows self-hosting and customization
•Category taxonomy organized around generative AI topics
•Built entirely through AI prompts showing rapid iteration

Weaknesses

•No technical differentiation from existing video platforms
•Most videos show zero to two views indicating no traction

Post Description

So I decided yesterday to do a full rewrite of the media gallery, turning it into a generative media catalog with full-suite social interactions, search, and personalized recommendations built on top of the samsar-js library. It took around 50 prompts in 5 sessions to get this done across the client and API, built with Codex GPT-5.6 Sol, Luna, and then Sol again (rate limit reset, thx Mr. Sama). I did not use screenshots for the UI, It was entirely hand-prompted end-to-end. The project is fully open-source under the MIT license. GitHub here: https://github.com/samsarone/Gallery

Similar Projects

AI/ML●●Solid

I solo built a text to AI motion graphics Video generator

Text-to-motion-graphics plus the ability to feed a YouTube snippet as inspiration is a practical, user-friendly twist that shortcuts idea-to-video. The prompt gallery and one-click MP4 export for vertical/horizontal formats show the builder focused on real creator workflows; the missing pieces are clarity on output quality, iteration controls, and model speed/cost — which will determine if this is genuinely useful beyond toy demos.

Ship ItSolve My ProblemSlick

ramsrigouthamg

304mo ago

SaaS●Mid