Jellyfish
Forget-C/JellyfishReviewed by AI
Checked by RepoPicks Radar on
What is Jellyfish?
Jellyfish is a self-hosted studio for short vertical dramas. You paste a script and it splits it into shots and pulls out characters, scenes, props, costumes and dialogue; you confirm each one, keep a shared library so the same character looks the same across shots, then start image and video jobs and follow them in a task centre. The picture and video making is done by AI providers you connect and pay for yourself. You run it with Docker Compose and open a web page on your own machine.
How easy is it to install Jellyfish?
D · terminal, several tools or a specific graphics card · Windows
- Windows D
- macOS D
- Linux D
All three platforms land on D through the same route: install Docker, copy the example .env, run "docker compose up --build" and wait for MySQL, Redis, a file store, the backend and the frontend to come up, then open http://localhost:7788 and add your own model providers. There is no installer and no ready binary in any release.
Coming soon: install in 1 click with the RepoPicks app.
Jellyfish on GitHub
- Stars
- 7,167on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +46stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Docker
- Disk space
- not determined
- Accounts or keys
- An AI model provider of your choice for text, image and video generation (set up inside the app)
- Cost
- Free, API costs extra · The software is free (Apache-2.0), but it does not generate anything on its own: you plug in your own AI providers for text, images and video and pay them per job. The repository gives no prices.
- License
- Apache-2.0 (Permissive)
- Can replace
- LTX Studio, AI
- Last change on GitHub
- Apr 20, 2026
Jellyfish risk signals (4)
- Attentiondevelopment appears stalled
The checks record the last commit on 2026-04-20, 198 days before this analysis, and the last release on 2026-04-17. The newest release is v0.3.2 and the repository was only created on 2026-03-06, so this is an early version that has not moved for months.
AI review
- Attentionrunning cost is on you and unstated
The README's capabilities are "multi-provider / multi-model management", "image and video generation tasks" and "single-shot and batch pre-checks": every shot of every episode is a paid call to an outside model. Nothing in the repository says which providers are supported or what a finished drama costs.
AI review
- Infofeature list reads as work in progress
The README hedges its own scope: "Preparation currently supports", "Current task infrastructure supports", "Provide a stable base for future AI workflow expansion". Some of the workflow is described as groundwork rather than finished.
AI review
- Infoseveral services published on your machine
The Docker Compose section lists the frontend on 7788, the backend on 8000, MySQL on 3306, Redis on 6379 and the RustFS file store on 9000, started from a copied .env.example. No login or access control for the web page is described in the README.
AI review
Evidence is shown in its original language.
What Radar could not check about Jellyfish
- Whether a script really comes out the other end as a finished video, since the repository shows only two screenshots and no sample output.
- Which AI providers actually work with it, and how much a short drama costs to generate.
- How much disk the database, uploads and generated media use (the checks could not estimate it).
- Whether the web page and prompts are fully translated, since the documentation is written in English, Japanese and Chinese in places.
- Whether the project is still maintained after months without commits.