Forget-C/Jellyfish

What it solves

Jellyfish is an end-to-end production workspace designed for AI-generated short dramas. It solves the fragmented nature of AI video production by combining script analysis, storyboarding, asset management, and video generation into a single, trackable workflow. It specifically addresses the problem of visual consistency (character, scene, and prop drift) across multiple shots in a series.

How it works

The system follows a structured production pipeline:

1. Script Breakdown: AI analyzes chapter scripts to extract shots, characters, scenes, props, costumes, and dialogue.
2. Shot Preparation: Users refine the extracted candidates, link them to existing assets, and mark shots as "ready."
3. Asset Management: A centralized entity model tracks characters, actors, scenes, and props to ensure the same visual identity is reused across different shots.
4. Generation Workspace: Users manage keyframes, reference images, and video prompts to launch image and video generation tasks.
5. Task Center: An asynchronous system tracks the status, progress, and results of text, image, and video generation jobs.

Who it’s for

• Short and micro-drama creators
• AI video studios producing content in batches
• Solo creators and brands producing story-driven promotional videos
• Education teams creating lesson videos

Highlights

• Integrated Workflow: Connects everything from the initial script to the final exported video.
• Consistency Control: Centralized management of characters and costumes to reduce visual drift.
• Async Task Tracking: A global task center for managing long-running AI generation jobs with cancellation and recovery.
• Extensible Infrastructure: Support for multi-provider model management and prompt templates.
• OpenAPI Integration: Automated frontend type generation based on the backend specification.