Getting Started with Toonflow — AI Short Drama Creation with Docker

🚀 How to Set Up Toonflow and Create Your First AI Short Drama in 30 Minutes

Alright, you've read the overview and you're intrigued. Now let's get our hands dirty.

I'll walk you through the actual setup — from Docker deployment to generating your first animated short drama. I've made all the mistakes so you don't have to. Let's go.

Prerequisites — What You'll Need

Before we start, gather these things:

• Docker installed on your machine (or a VPS with Docker)
• An LLM API key — Toonflow works with OpenAI, Claude, or any OpenAI-compatible provider
• A video model API — options include Seedance, Doubao, or any Sora-compatible API
• An image model API — GPT Image 2 or any OpenAI-compatible image API work
• About 4GB of RAM minimum (8GB recommended for smooth operation)

Heads up: I spent my first 30 minutes wondering why nothing was generating. Turned out I forgot to set the API keys. Don't be like me — have them ready before you start.

Step 1: Deploy with Docker

This takes about 30 seconds. Just run:

docker pull xkrfer/toonflow-app:latest
docker run -d \
  --name toonflow \
  -p 8080:8080 \
  -v ./toonflow_data:/data \
  xkrfer/toonflow-app:latest

I chose port 8080 to avoid conflicts with anything else running. If 8080 is taken, change the left side (e.g., -p 9090:8080).

👉 If you see the container starting without errors — mazel tov, you're in. If not, check docker logs toonflow to see what went wrong. 9 times out of 10 it's a port conflict.

Step 2: Login and Configure

Open http://localhost:8080 in your browser. Login with the default credentials:

Username: admin
Password: admin123

First thing you should do: Go to the Settings Center and configure your model providers. This is where I lost 20 minutes — Toonflow won't do anything until you connect at least one LLM provider and one video model.

Step 3: Create Your First Project

Now the fun part:

1. Click "New Project" and give it a name
2. Import your source material — paste a novel, a script, or even just a paragraph describing your idea
3. Run Chapter Event Extraction — Toonflow will automatically break your story into chapters and extract key events. This took about 15 seconds on my test with a 3-chapter short story
4. Enter ScriptAgent — the agent generates the story skeleton, adaptation strategy, and structured script. This is where the three-layer agent system shines — the supervision agent actually caught a character inconsistency in my draft and flagged it
5. Switch to ProductionAgent — organize storyboards, characters, and video nodes on the infinite canvas

Step 4: Generate and Export

Once your storyboard looks good:

• Refine storyboard images — Toonflow lets you tweak each frame before generating video
• Send to video generation — each storyboard segment gets turned into video clips
• Stitch and export — the final output is a complete short drama

The demo on the repo took about 2 hours to produce a 2-minute clip (including generating ~¥120 worth of video models). Your mileage will vary depending on the models you use and the length of your content.

Common Pitfalls (I Made All of These)

Final Thoughts

Toonflow is still early-stage — the UI is functional rather than polished, and you'll need to bring your own API keys for the heavy lifting models. But the workflow design is genuinely smart. The infinite canvas, the agent collaboration system, the persistent memory — these are features I haven't seen combined in any other open-source video creation tool.

If you're a content creator, a novelist exploring visual adaptation, or just an AI enthusiast who wants to see what the bleeding edge looks like — give it a spin. The Docker setup takes 2 minutes, and you could have your first AI short drama in an afternoon.