Case study

Pulse (Micro-Drama Platform)

A vertical short-form micro-drama platform (TikTok-style feed) paired with an AI-powered Story Lab for building serialized stories end-to-end: idea intake → story bible → beat sheets → scripts → exports and creative assets.

Visit live site

Problem

  • A vertical short-form feed requires a smooth full-screen player UX and strong content hierarchy (Series → Episodes → Videos).
  • Collaborative writing needs teams, permissions, and a durable revision model (scripts, rewrites, and history).
  • AI generation workflows span multiple steps (bible → beats → scripts → rewrites) and should run asynchronously with progress/status.
  • Realtime collaboration and status updates benefit from WebSockets and a Redis-backed channels layer.

Solution

  • Built the viewing experience as a snap-style vertical feed with a full-screen player and series/episode organization.
  • Designed Story Lab around structured intakes (JSON schema) to standardize discovery and reduce prompt drift.
  • Implemented an AI generation pipeline for bibles, beat sheets, scripts, and rewrites with revision tracking and typed script blocks.
  • Used Celery + Redis for async jobs, integrating OpenRouter (LLM) and Replicate (image generation) behind clear provider boundaries.
  • Added Django Channels for realtime updates, backed by `channels-redis`.
  • Containerized the stack for local development and deployment (Postgres, Redis, Nginx, API, workers).

Architecture

Results

Vertical feed UX

Full-screen viewing experience optimized for short-form content

Async AI pipeline

Multi-step generation runs in background jobs with progress/status

Collaborative writing

Teams, roles, and revisioned script artifacts

Export-ready

Scripts and production artifacts export to industry formats

How this maps to client use cases

  • Any product that needs multi-step AI workflows can use the same async pipeline approach to keep UX responsive.
  • Teams can standardize generation quality using structured intake templates and typed content models.
  • Realtime status/progress and collaborative editing patterns apply to many creator tools and internal platforms.