Install the skill from ClawHub
Start from the PixelDojo skill listing so OpenClaw gets the packaged scripts, setup instructions, and reference files.
This page is the OpenClaw-specific companion to PixelDojo AI Agents. It focuses on installing the PixelDojo skill from ClawHub, exporting `PIXELDOJO_API_KEY`, checking the live catalog, and running the exact OpenClaw commands that create images or videos.
Workflow
The install flow is simple: install the skill, export the key, inspect the live catalog, then generate and poll the async job.
Start from the PixelDojo skill listing so OpenClaw gets the packaged scripts, setup instructions, and reference files.
Set PIXELDOJO_API_KEY before running helpers so every request is authenticated against the live PixelDojo API.
Run models.sh first, choose the best fit for the request, and do not guess model IDs from memory.
Use generate.sh or the Nano Banana helper, poll the async job, and return the saved asset path to the user.
Commands
These examples match the current PixelDojo skill workflow on ClawHub: setup, catalog lookup, image generation, video generation, Nano Banana, and direct status polling.
Export your key and optionally copy the local env template.
export PIXELDOJO_API_KEY=your_api_key_here
cp ~/.openclaw/skills/pixeldojo/.env.example ~/.openclaw/skills/pixeldojo/.envAlways inspect the live catalog first so OpenClaw uses current model IDs and capabilities.
bash ~/.openclaw/skills/pixeldojo/models.shUse the shared prompt-based helper for standard text-to-image generation.
bash ~/.openclaw/skills/pixeldojo/generate.sh image "editorial product photo of a silver robot" flux-2-pro --aspect-ratio 16:9Use the video mode when the request needs motion, duration, or image-to-video input.
bash ~/.openclaw/skills/pixeldojo/generate.sh video "cinematic ocean waves at sunset" seedance-1.5 --duration 5
bash ~/.openclaw/skills/pixeldojo/generate.sh video "slow camera push-in" wan-2.6-flash --image-url https://example.com/input.png --duration 5Reach for the dedicated helper when prompt adherence or typography quality matters most.
python3 ~/.openclaw/skills/pixeldojo/scripts/generate-nano-banana.py "clean ecommerce hero shot of running shoes" --aspect-ratio 16:9 --output ~/Desktop/shoes.pngPoll the async job directly when you already have a job ID and want the latest status.
bash ~/.openclaw/skills/pixeldojo/status.sh job_abc123Model picks
If the user asks for a default recommendation, these are good starting points. If the request is specialized, check the live catalog before choosing.
Strong default pick for polished image generation tasks.
flux-2-maxBest fit when the user cares about text rendering or exact instructions.
nano-banana-2Use for editing flows and context-sensitive image transformations.
flux-kontext-proA strong OpenClaw option for quick motion from a reference image.
wan-2.6-flashGood first stop for short cinematic video prompts.
seedance-1.5FAQ
Open the ClawHub listing, install the PixelDojo skill into OpenClaw, then export PIXELDOJO_API_KEY before running any helper scripts.
The required variable is PIXELDOJO_API_KEY. You can also override the API base with PIXELDOJO_API_BASE if you need a non-default endpoint.
Run models.sh first so you are selecting from the live PixelDojo catalog instead of guessing model IDs or outdated capabilities.
Yes. The PixelDojo skill supports image generation, text-to-video, image-to-video, async job polling, and direct downloads of the completed output.
Use the dedicated Nano Banana helper when the user specifically wants Nano Banana 2 or when prompt adherence and typography quality are especially important.