Jellyfish

Jellyfish — AI Short Drama Studio

  English ·
  日本語

An end-to-end production workspace for AI-generated short dramas.  
From script input to structured storyboarding, consistency management,
shot preparation, video generation, and export.

📷 Screenshots

| Project overview | Asset management |

✨ Core Value

- Connect the full production flow: Move from script input to storyboard preparation, image/video generation, and task tracking in one place.
- Turn AI output into reusable production assets: Shots, candidate assets, dialogue, prompts, and generation tasks can all be reviewed and reused.
- Treat consistency as a first-class problem: Centralized character, scene, prop, and costume management reduces drift across shots.
- Handle long-running generation as trackable tasks: Text, image, and video jobs all go through one async task system with status, cancel, and recovery.
- Build AI capability as infrastructure: Model management, prompt templates, files, and OpenAPI-based collaboration make the system extensible.

✨ Core Capabilities

Jellyfish is not just a single “AI image/video” utility. It is a
production workspace built around:

- script understanding
- shot preparation
- asset consistency
- generation execution
- task tracking

1. AI script understanding and storyboard breakdown

- Split chapter scripts into shots
- Extract characters, scenes, props, costumes, and dialogue
- Run script optimization, simplification, and consistency checks
- Support targeted analysis such as character portraits or scene details

2. Shot preparation and confirmation workflow

The main workflow is:

script breakdown → shot preparation → candidate confirmation → shot ready → generation workspace

Preparation currently supports:

- extracting and refreshing shot candidates
- accepting or ignoring asset candidates
- accepting or ignoring dialogue candidates
- linking existing characters, scenes, props, and costumes
- correcting shot-level basic information
- using a unified readiness state to decide whether a shot is prepared

3. Asset consistency and reuse

The system maintains a shared entity model across:

- characters / actors
- scenes
- props
- costumes

This supports asset reuse across shots and helps stabilize style and identity.

4. Shot-level image and video orchestration

Once a shot is ready, the generation workspace supports:

- keyframe and reference image management
- shot-level video prompt preview
- image and video generation tasks
- single-shot and batch pre-checks
- writing generation outputs back into the shot/media system

5. Unified async task center

Current task infrastructure supports:

- async text-processing tasks
- async image and video generation tasks
- unified task status, result, and elapsed-time tracking
- task cancellation
- a global task center with context-aware navigation back to project/chapter/shot

6. Model, prompt, and generation infrastructure

Supporting capabilities include:

- multi-provider / multi-model management
- default model settings by category
- prompt template management
- file and generated media management
- OpenAPI-driven frontend/backend contracts

🚀 Feature Overview

Project and chapter management

……（后略，见原文）