Dream-to-Video Skill

中文文档

AI agent skill that transforms dream text into cinematic videos. Describe your dream, and the agent automatically generates professional video prompts, submits them to the Jimeng platform via browser automation, downloads the finished videos, and applies post-processing effects.

How It Works

Dream text → AI transforms to cinematic prompt → Queue → Browser automation → Jimeng platform → Download → Post-processing → Video files

1. Prompt transformation — The AI agent converts your dream description into a detailed cinematic video prompt following 10 strict rules (photorealistic style, fisheye lens, no text overlays, silent narrative, etc.)
2. Queue submission — The prompt is added to a local task queue (SQLite + JSONL)
3. Browser automation — A background worker drives Chromium via Playwright to interact with the Jimeng video generation platform: login, submit prompts, upload reference images when needed, monitor progress
4. Download & post-processing — Completed videos are downloaded and automatically processed with an "Elliptic Shatter" edge effect (center-clear, shattered-glass edges), outputting both the original and the effect version

Install

Via skills CLI

npx skills add mediastormDev/dream-to-video-skill -s dream-to-video

Manual install

Clone the repo and symlink the skill directory into your agent's skill folder:

git clone https://github.com/mediastormDev/dream-to-video-skill.git
mkdir -p ~/.claude/skills
ln -s "$(pwd)/dream-to-video-skill/skills/dream-to-video" ~/.claude/skills/dream-to-video

Requirements

• Python >= 3.10
• Chromium (installed via Playwright)
• An API key from any supported provider:

Provider	Model	Get a key
Claude (default)	claude-sonnet-4	console.anthropic.com
OpenAI	gpt-4o	platform.openai.com
OpenRouter	any model	openrouter.ai
Google Gemini	gemini-2.5-flash	aistudio.google.com

Setup

After installing the skill, clone this repo to get the Python toolchain:

git clone https://github.com/mediastormDev/dream-to-video-skill.git
cd dream-to-video-skill/dream_to_video
pip install -r requirements.txt
playwright install chromium

Then log in to the Jimeng platform (one-time QR code scan):

python main.py login

Usage

Once set up, just describe a dream to your AI agent:

"I dreamed I was running barefoot on a coastal boardwalk surrounded by hundreds of sea lions..."

The agent will:

1. Transform your text into a cinematic video prompt
2. Submit it to the queue (python main.py add "<prompt>")
3. Start the background worker (python main.py worker)
4. Notify you when the video is ready in dream_to_video/output/

CLI Commands

Command	Description
python main.py login	Log in to Jimeng via QR code
python main.py verify	Check login status
python main.py add "<prompt>"	Add a prompt to the task queue
python main.py worker	Start background worker
python main.py status	View task progress
python main.py generate "<prompt>"	Single synchronous generation
python main.py serve	Start FastAPI web server

Output

Each task produces two video files in dream_to_video/output/:

• task_XXX_YYYYMMDD_HHMMSS.mp4 — Original video
• task_XXX_YYYYMMDD_HHMMSS_elliptic-shatter.mp4 — With shattered-glass edge effect

License

MIT