An AI film studio that runs on your laptop (and a pile of API keys)

What it does

LumenX Studio is a browser-based production suite that reads a text script, uses an LLM to extract characters and scenes, generates matching art assets and storyboards, and stitches everything into a voiced video with FFmpeg. A separate Playground module offers standalone image and video generation across ten-plus models. It is essentially a local project manager for a fleet of remote AI renderers.

The interesting bit

The real work here is plumbing: the project normalizes APIs from DashScope, Kling, Vidu, MuleRun and others behind a unified Next.js frontend and Python pipeline, then adds comic-specific workflow steps like character turnarounds and automated dubbing. The “local-first” label means your files stay on disk, but every pixel is still rendered in the cloud.

Key highlights

• End-to-end pipeline from script analysis to export: LLM parsing, art-direction controls, I2V/R2V generation, TTS dubbing, and timeline composition.
• Playground mode for ad-hoc creation with 10+ models including Wan 2.7, Kling V3, Seedance 2.0, and GPT-Image-2.
• Tiered configuration: a single DashScope key gets you started; adding MuleRun, Kling, or Vidu credentials unlocks extra models.
• Model catalog is driven by YAML configs, so new providers can be onboarded without touching core code (according to the docs).

Caveats

• The README is entirely in Chinese; the English translation appears to be a stub or secondary.
• “Local-first” refers to storage and orchestration, not inference: every generative feature still requires external paid API keys.

Verdict

Worth a look if you’re building narrative video tools or need a reference architecture for multi-model AI pipelines. If you want fully self-hosted diffusion weights, this is just a very capable remote-client.