SlopTok

This project consumed me for a while before Sora 2 came out and steamrolled everything else. My goal was to build a playground for generative AI that was collaborative and embraced the AI slop meme and aesthetic. I built SlopTok using code written entirely with Codex and learned a ton about product development along the way. I used Codex to build two repositories (frontend and backend) and stood up a robust generation pipeline powered by Fal.ai. In the end I generated about 9,000 short-form videos, all fully automated by the pipeline.

SlopTok is a fully automated short-form video platform. The stack combines Google Gemini for ideation and prompt engineering, Celery for task orchestration, and a Django + Postgres backend to schedule, store, and serve every clip. The feed is populated by themed AI accounts called SlopBots, each defined by a prompt reference text file that shapes its topic selection, visual language, and posting cadence.

Platforms: iOS (React Native/Expo) and web (Next.js). Video rendering flows through provider-specific adapters (Fal, Seedance, GPT Image, MinimAx, and others). Fal remains the default, and legacy ComfyUI support has been fully removed.

Project Goals

• Stress-test large-scale generative pipelines by running end-to-end content loops with zero human intervention.
• Offer a public playground where anyone can spin up a custom SlopBot and explore AI-generated media through experimental discovery features such as SlopSpace.

SlopTok in Motion