Skip to content
English

Skills

After HiAPI Skills are installed, submit image or video generation requests to Claude Code, Codex, or OpenClaw. Generation skills apply the documented model parameters, start the generation, and return the result. Prompt-only skills prepare or review prompts without creating an API task.

Generation skills use the Unified Async API and require HIAPI_API_KEY. Prompt-only skills do not create API tasks and do not require an API key.

Each card provides an installation command. Copy the command and execute it in a terminal.

GPT Image 2

Generate images from text, or edit and combine images with reference inputs. The available standard, Pro, and image-to-image variants provide different quality and cost tiers for posters, illustrations, product shots, and covers.

Model parameters · GitHub

npx -y github:HiAPIAI/hiapi-gpt-image-2-skill -y

Seedream 5.0 Pro

Generate and edit images with accurate in-image text, brush calligraphy, photorealistic detail, and multi-reference composition. Use quality=basic for 1K iteration or quality=high for 2K delivery.

Model parameters · GitHub

npx -y github:HiAPIAI/hiapi-seedream-5-0-pro-skill -y

Seedance 2.5

Produce text-to-video, image-to-video, and reference-to-video clips up to 30 seconds with native audio. Includes preflight estimates, idempotent submission, task recovery, downloads, and technical and creative quality checks.

Model parameters · GitHub

Seedance 2.0

Flagship video generation with text-to-video, first/last-frame control, image/video/audio references, and synchronized audio. Suitable for cinematic video production.

Model parameters · GitHub

HappyHorse 1.0

Fast text-to-video generation for 3–15 second clips at up to 1080p. Suitable for short clips, rapid validation, and cost-sensitive workloads.

Model parameters · GitHub

Video Prompt Generator

Expands a brief idea, link, or topic into a complete video prompt with shot descriptions, camera movement, on-screen text, and pacing for use with the video skills above.

GitHub
npx -y github:HiAPIAI/hiapi-video-prompt-generator-skill -y
View the HiAPI Skills collection
  1. Set up your API key

    Create one at Dashboard → API Keys, then export it (Node.js 18+ required):

    Terminal window
    export HIAPI_API_KEY="your_hiapi_api_key_here"
  2. Install the skills

    The commands below install all six listed skills; each installer auto-detects Claude Code / Codex and selects the appropriate directory:

    Terminal window
    npx -y github:HiAPIAI/hiapi-gpt-image-2-skill -y
    npx -y github:HiAPIAI/hiapi-seedream-5-0-pro-skill -y
    npx -y github:HiAPIAI/hiapi-seedance-2-5-video-skill -y
    npx -y github:HiAPIAI/hiapi-seedance-2-0-video-skill -y
    npx -y github:HiAPIAI/hiapi-happyhorse-1-0-video-skill -y
    npx -y github:HiAPIAI/hiapi-video-prompt-generator-skill -y
  3. Start using the skills

    After installation, submit requests to your agent. For example:

    “Generate a minimal product poster with GPT Image 2”

    “Turn this idea into a video prompt first, then generate it with Seedance”

    “Make a 3-second test video with HappyHorse”

Results are downloaded to a local outputs/ directory when supported. Remote CDN links expire after about 7 days; download files that must be retained.

Before each run, skills check their version against the official skills.json catalog. A newer version produces an upgrade notice; breaking API changes require the installation command to be run again. The check is skipped when offline and does not block generation.

Set HIAPI_SKIP_UPDATE_CHECK=1 to turn the check off.

SymptomFix
HIAPI_API_KEY is requiredExport the key — create one at Dashboard → API Keys
HTTP 401 / 403Verify the API key is correct and active
HTTP 402, insufficient balanceAdd funds at Dashboard → Billing; see pricing
HTTP 400 parameter errorsCheck the parameter combination on the model pages
HTTP 429Wait and retry, or reduce concurrent generations
”Update required” noticeRe-run the install command, then retry
Waiting times outThe task may still be generating — check later in the dashboard or via GET /v1/tasks/:id

For a single integration covering multiple models, use Remote MCP. For product integration, call the Unified Async API directly.