A full film crew in Python: AI agents script, storyboard, and shoot
It exists because raw text-to-video models don't understand story arcs, and manually hopping between scripting, image, and video tools to compensate is tedious.

What it does
MicroDrama AI is a FastAPI and Next.js application that automates short-form video production from a text prompt or existing script. It routes creative tasks through a chain of LLM agents—screenwriter, character extractor, and storyboard artist—then triggers image generation, image-to-video conversion via MuAPI, and finally concatenates 5-second clips into a single MP4. The frontend streams each stage in real time over SSE, turning a fragmented manual workflow into a single-button pipeline.
The interesting bit
The project is essentially a film production crew implemented as Python functions: creative pre-production happens via LLM agents talking to MuAPI’s Claude endpoint, while visual execution is delegated to Flux and Kling through the same unified key. It is, kindly put, mostly glue code—well-organized glue that replaces a human producer frantically tabbing between separate scripting, image, and video tools.
Key highlights
- Two entry modes: drop a raw idea for full autopilot, or supply a script to skip the story-development phase.
- Character consistency is enforced by extracting static and dynamic visual features, generating reference portraits, and feeding them into downstream image-to-image and image-to-video stages.
- Real-time SSE streaming exposes the pipeline state with an animated stage timeline, so you can watch the agents work.
- The README emphasizes that a single MuAPI key powers the entire flow, abstracting LLM, image, and video inference behind one account.
Caveats
- The repository description name-checks Seedance 2, Veo 3.1, and Sora 2, but the README architecture diagram only documents a MuAPI pipeline using Flux for images and Kling v2.1 Standard for video.
- Because nearly all inference happens through MuAPI’s paid endpoints, the project functions as an orchestration layer rather than a self-contained engine.
Verdict
Ideal for creators who want to batch-produce short-form dramas without juggling multiple AI tool subscriptions; less useful if you are looking to run local diffusion models or avoid third-party API billing.
Frequently asked
- What is Anil-matcha/Open-AI-Micro-Drama-Generator?
- It exists because raw text-to-video models don't understand story arcs, and manually hopping between scripting, image, and video tools to compensate is tedious.
- Is Open-AI-Micro-Drama-Generator open source?
- Yes — Anil-matcha/Open-AI-Micro-Drama-Generator is an open-source project tracked on heatdrop.
- What language is Open-AI-Micro-Drama-Generator written in?
- Anil-matcha/Open-AI-Micro-Drama-Generator is primarily written in Python.
- How popular is Open-AI-Micro-Drama-Generator?
- Anil-matcha/Open-AI-Micro-Drama-Generator has 500 stars on GitHub.
- Where can I find Open-AI-Micro-Drama-Generator?
- Anil-matcha/Open-AI-Micro-Drama-Generator is on GitHub at https://github.com/Anil-matcha/Open-AI-Micro-Drama-Generator.