MicroDrama AI Generator
Turn a premise or script into storyboarded scenes and a stitched short drama through one automated pipeline.
- Source repo
- Anil-matcha/Open-AI-Micro-Drama-Generator
- Stars
- ★ 530
- Last updated
- 2mo ago
- Primary language
- Python
- FA score
- 43/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specific
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A short-form video creator wants to expand a premise into scenes and generate clips for the resulting shots.
- Not a fit if
- Teams that need to switch model providers
- Creators who need generation to stay local and avoid uploading assets
- Source review
- 43/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
MicroDrama AI is a short-drama generation app with a FastAPI server and a Next.js 14 web client. Users submit either an idea or an existing script, and the server runs screenwriting, character extraction, and storyboard agents before generating portraits, shot frames, and video clips through MuAPI. MoviePy joins the clips into an MP4, while the web client streams pipeline progress over SSE and offers playback and download. The workflow requires a MuAPI API key and uses its documented LLM, image, and video endpoints, so it fits creators willing to depend on that provider and cloud generation.
The client accepts an idea or a script-mode submission; POST /api/generate starts a job and returns a job_id. In idea2video, Screenwriter develops a story and writes 2–4 scenes, CharacterExtractor extracts characters, and StoryboardArtist designs 3–5 shots per scene; script mode skips story development and scene writing. The server calls MuAPI's claude-sonnet-4-6, flux-dev-image, flux-kontext-dev-i2i, and kling-v2.1-standard-i2v endpoints to produce text, portraits, first frames, and 5-second clips per shot, then MoviePy concatenates them into final_video.mp4. The client consumes SSE from GET /api/status/{job_id} for progress and plays or downloads the completed video.
- A short-form video creator wants to expand a premise into scenes and generate clips for the resulting shots.
- An independent filmmaker with a scene script wants to skip writing and generate character references, a storyboard, and video.
- A creator needs recurring character appearance across shots and can use generated portraits as visual references.
- A small production team wants a web interface that shows progress from story development through final concatenation.
- A developer wants to self-host a FastAPI and Next.js drama-generation app that calls generation services with one MuAPI key.
How do you install or deploy this agent?
Requires Python 3.10+, Node.js 18+, and a MuAPI API key. Install and start the server and client separately:
git clone https://github.com/Anil-matcha/Open-AI-Micro-Drama-Generator.git
cd Open-AI-Micro-Drama-Generatorcd server
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .envSet MUAPI_KEY in server/.env, then start FastAPI:
uvicorn api:app --reload --port 8000In another terminal, start the Next.js client:
cd client
npm install
npm run devHow do you use this agent?
Open http://localhost:3000, enter an idea or switch to script mode and provide a script; optionally add requirements and choose a visual style, then click Generate Video. The API can also start a job directly:
curl -X POST http://localhost:8000/api/generate \
-H 'Content-Type: application/json' \
-d '{"idea":"A detective chases a shadow through a neon-lit cyberpunk city","user_requirement":"Keep it under 3 scenes, suitable for all ages","style":"Cinematic","mode":"idea2video"}'Use the returned job_id to subscribe to SSE progress at GET /api/status/{job_id}. The completion event includes the video path; the client displays the result and provides a download control.
What are this agent's strengths and limitations?
- Supports both idea-to-video and script-to-video; script mode skips story development and scene writing.
- Extracts characters, generates reference portraits, and uses them for shot frames to help preserve visual identity across shots.
- Streams stage progress over SSE and replays prior events after a reconnect.
- Uses one MuAPI key for the documented LLM, image, and video generation endpoints.
- The generation pipeline depends on MuAPI's LLM, image, and video endpoints; the README documents no provider adapter for switching services.
- Generated clips are 5 seconds per shot, and the README does not describe extending that duration or configuring broader shot limits.
- Generation requires network access and MuAPI quota; the documented error event includes quota exhaustion.
- The README gives no actual MuAPI model prices or per-video cost estimate.
How does this agent compare with similar options?
The README describes raw text-to-video as lacking story structure, scene continuity, and character consistency; this project adds screenwriting, character extraction, storyboarding, reference images, and clip concatenation. It provides no direct feature comparison with another drama generator.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| MicroDrama AI Generator This agent | 43 · Major gaps | Web appFree + model costs | ★ 530 | 2mo ago | Python | — |
| ArcReel Video Creation Workspace | 52 · Major gaps | Self-hosted serviceFree + model costs | ★ 5.5k | today | Python | OpenAI API · Claude API |
| waoowaoo AI Film Studio | 0 · Major gaps | Self-hosted serviceFree + model costs | ★ 15k | 22d ago | TypeScript | — |
| Generative Media Skills | 41 · Major gaps | CLIFree + model costs | ★ 5.8k | today | Python | Claude Code · Claude.ai |
How does FollowAgents rate this agent?
Why each dimension lost points
The README says one MUAPI_KEY is used for LLM, image, and video generation and file uploads, making the main data flow reasonably visible (data_flow_transparency=2). It does not explain key protection or how inputs and generated content are retained or handled; least privilege and sensitive-data handling are only thinly documented (least_privilege=1, sensitive_data_handling=1). A submitted generation request calls external models, uploads images, and generates video, but the README gives little detail on confirmation, controlling external effects, or recovery and rollback (user_confirmation=1, external_effects=1, rollback=0). It lists MuAPI, MoviePy, and installation steps but provides no evidence of pinned versions or dependency security practices (dependency_security=1). The README claims MIT and links to a LICENSE file, but the supplied material does not include the license text; authorship and a maintenance contact are also unclear (source_attribution=1).
The README documents pipeline stages, SSE events, reconnection behavior, and an example quota error, giving failure reporting some support (failure_messages=2). However, the description names Seedance 2 VIP, Kling 3.0 Pro, Veo 3.1, and Sora 2, while the README lists models including Kling v2.1; this weakens consistency (self_consistency=1). The pipeline depends on MuAPI, external model services, and polling, with no availability assurance or offline alternative described (dependency_availability=1).
The README covers idea-to-video and script-to-video, and gives style choices, optional requirements, and an age-appropriateness example, providing reasonable coverage of ordinary scenarios (audience_and_scenarios=2). Modes, input fields, and pipeline stages are clearly described (trigger_precision=2), but model capability, content, and failure boundaries are only partly covered (capability_boundaries=1). Python, Node.js, ports, environment variables, and startup steps are documented for a common local development setup (environment_fit=2).
The README provides a directory map, architecture, pipeline, API, configuration, and UI overview, with strong information organization (information_architecture=3). Setup steps and request examples are useful, but dependency locking, deployment, and troubleshooting details are absent (install_notes=2). Project and model naming is inconsistent between the description and README (naming_stability=1). The README includes usage steps and examples for requests and responses, but no FAQ (examples_and_faq=2). Known limitations are not adequately documented (known_limitations=1). MIT is claimed in the README, but the supplied material does not include the license text (license=1). No version history, changelog, named maintainer, or update path is shown (versioning_changelog=0, maintenance_responsibility=0).
The README describes a complete path from idea or script through storyboarding, image and video generation, and MP4 assembly, with a download flow, giving the output practical usability (output_usability=2). Combining writing, character reference images, and video generation offers potential value over a fragmented manual workflow, but the claimed results are not corroborated in the supplied material (marginal_value=2). The README says a paid API key is required but gives no cost, quota, or per-generation estimate (cost_benefit=1).
The README names API routes, models, stages, and event formats, making some claims traceable to specific interface descriptions, but implementation files are not supplied for verification (claim_traceability=1). Only the README is available, so claims cannot be corroborated across sources (cross_source_corroboration=0). Setup and API details are presented alongside product claims without clearly separating implemented facts from promotional outcome claims (fact_inference_separation=1).
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Model names differ between the description and README; verify supported models and versions against the code and current MuAPI interfaces.
- Before use, clarify how the MuAPI key is protected, how uploaded files and generated content are handled and retained, and what API calls cost. This review did not run the code or verify outputs.
- The supplied material does not include the license text, a dependency lockfile, security maintenance details, or a named maintainer.
FAQ
What credentials does generation need?
MUAPI_KEY in server/.env. The README says the same key is used for LLM, image, and video generation.Can I start from an existing script?
script2video and provide a script; the pipeline starts with character extraction and skips story development and scene writing.How is the finished video delivered?
final_video.mp4 with MoviePy. The web client provides playback and download controls, and the API completion event includes a video path.Does it show why a generation failed?
What license applies?
LICENSE file; the repository metadata supplied here says the license is unknown, so verify the license file before adopting it.