AgentSkillsCN

fal-ai

利用 fal.ai 的模型库生成图像、视频、音乐等多种内容。支持 100 多种模型，包括 Flux、Stable Diffusion、视频生成、超分辨率增强以及音频生成。

SKILL.md
--- frontmatter
name: fal-ai
description: Generate images, videos, music, and more using fal.ai's model library. Supports 100+ models including Flux, Stable Diffusion, video generation, upscaling, and audio.
homepage: https://fal.ai
metadata: {"openclaw":{"emoji":"🎨","requires":{"env":["FAL_KEY"]}}}

fal-ai

Universal wrapper for fal.ai — access 100+ AI models for image generation, video, music, upscaling, and more.

Setup

  1. •Get your API key at https://fal.ai/dashboard/keys
  2. •Set the environment variable:
    bash
    export FAL_KEY="your-key-here"

Usage

bash
# Generate an image
python3 fal.py generate "fal-ai/flux/schnell" '{"prompt": "a robot reading a book"}'

# Generate music
python3 fal.py generate "fal-ai/minimax-music/v2" '{"prompt": "upbeat electronic track"}'

# Upscale a video
python3 fal.py generate-queue "fal-ai/flashvsr/upscale/video" '{"video_url": "https://..."}'

# Search for models
python3 fal.py search "video generation"

# Get model schema (see required parameters)
python3 fal.py schema "fal-ai/flux/schnell"

Popular Models

Image Generation

ModelCostNotes
fal-ai/flux/schnell$0.003Fast, great quality
fal-ai/flux-pro/v1.1$0.05Best quality
fal-ai/nano-banana-pro$0.15Google Imagen 3, great typography
fal-ai/stable-diffusion-v3-medium$0.035SD3

Video Generation

ModelCostNotes
fal-ai/minimax-video/video-01-live~$0.50MiniMax video
fal-ai/kling-video/v1.5/pro~$0.30Kling video
fal-ai/luma-dream-machine~$0.30Luma Labs

Video Upscaling

ModelCostNotes
fal-ai/flashvsr/upscale/video$0.0005/MPFast upscale
fal-ai/bytedance-upscaler/upscale/video~$0.007/sByteDance, up to 4K

Music & Audio

ModelCostNotes
fal-ai/elevenlabs/musicvariesElevenLabs text-to-music
fal-ai/minimax-music/v2variesMiniMax music generation

Commands

CommandDescription
generate <model> <json>Run model directly (fast, small jobs)
generate-queue <model> <json>Run via queue (longer jobs)
modelsList all available models
search <keywords>Search for models
schema <model>Get model's input schema
status <url>Check queued job status
result <url>Get completed job result
cancel <url>Cancel queued job

Examples

Generate an image and save it

bash
# Generate
python3 fal.py generate "fal-ai/flux/schnell" '{"prompt": "cyberpunk cityscape at night"}' > result.json

# Extract URL (the response includes image URLs)
cat result.json | jq -r '.images[0].url'

Queue a long video job

bash
# Submit to queue
python3 fal.py generate-queue "fal-ai/minimax-video/video-01-live" \
  '{"prompt": "a cat playing piano"}' > queue.json

# Check status
python3 fal.py status "$(cat queue.json | jq -r '.status_url')"

# Get result when done
python3 fal.py result "$(cat queue.json | jq -r '.response_url')"

Reference image (image-to-image)

bash
python3 fal.py generate "fal-ai/flux/dev/image-to-image" '{
  "prompt": "make it look like a watercolor painting",
  "image_url": "https://example.com/photo.jpg",
  "strength": 0.7
}'

Tips

  • •Use generate for fast models (< 30s)
  • •Use generate-queue for slow models (video, large images)
  • •Check schema to see required/optional parameters
  • •Most image models return images[0].url in the response
  • •Video models typically return video.url

Links