v0.2.0 — First Public Release

Phantom Canvas v0.2.0

Anti-detection browser wrapped as HTTP API for AI image/video generation via Gemini Web. No API keys needed.

Features

• Text-to-image — send a prompt, get a generated image
• Image-to-image — upload reference images for consistent character design
• Multi-turn conversation — iterate on designs in the same Gemini chat
• Video generation — walk cycle animations via Gemini's Veo integration
• Async task queue — POST /generate returns immediately, poll or use webhooks
• CLI — phantom-canvas login / phantom-canvas (headless server)
• Session management — auto-detection, expiry prompts, stored in ~/.phantom-canvas/
• Multi-locale — works with English, Danish, German, and other Gemini UI languages

Install

bun install -g github:baixianger/phantom-canvas
phantom-canvas login   # first time: login to Google
phantom-canvas         # start API server on :8420

Quick Example

# Generate
curl -X POST localhost:8420/generate \
  -d '{"prompt": "isometric pixel art knight, FFT style, green background"}'

# Check result
curl localhost:8420/task/{task_id}

# Download image
curl localhost:8420/task/{task_id}/image/0 -o sprite.png

See examples.md for full usage guide including multi-turn, video, webhook, and pipeline examples.

Built With

• Bun — runtime
• camoufox-js — anti-detection browser
• Hono — HTTP framework