fal-ai-model-search

Original：🇺🇸 English

Translated

Search and integrate Fal AI models from fal.ai platform. Use when the user wants to (1) search for models on Fal AI platform, (2) get detailed information about a specific Fal AI model, (3) integrate a Fal AI model into the project, (4) explore available AI models on fal.ai, or mentions "fal.ai", "图像生成", "AI video model", "text to image", "text-to-speech".

2installs

Sourcesweetretry/skills

Added on2026-02-07

NPX Install

npx skill4agent add sweetretry/skills fal-ai-model-search

SKILL.md Content

View Translation Comparison →

Fal AI Model Search & Integration

Guide for searching and integrating Fal AI models from fal.ai platform.

Workflow

1. Search Models

When user wants to find a model, use WebFetch to search:

https://fal.ai/explore/search?q=<search-query>

Process:

Parse search results - look for model IDs in format

<namespace>/<model-name>

(e.g.,

fal-ai/flux-schnell

,

xai/grok-imagine-image

)

Extract: model ID, category (text-to-image, image-to-image, etc.), description
Present the found models to the user with:
- Model name and category
- Description
- Playground link:
```
https://fal.ai/models/<full-model-id>
```
  (for xai:
```
https://fal.ai/models/xai/<model>
```
  ; others:
```
https://fal.ai/models/fal-ai/<model>
```
  )
Wait for user confirmation on which model to use

Important: The full Model ID format depends on the model namespace:

Search Result Full Model ID API Endpoint

Search Result	Full Model ID	API Endpoint
`xai/grok-imagine-image`	`xai/grok-imagine-image`	`https://fal.run/xai/grok-imagine-image`
`bytedance/seedream/v4.5/...`	`fal-ai/bytedance/seedream/v4.5/...`	`https://fal.run/fal-ai/bytedance/seedream/v4.5/...`
`flux-1/schnell`	`fal-ai/flux-1/schnell`	`https://fal.run/fal-ai/flux-1/schnell`

xai/grok-imagine-image

xai/grok-imagine-image

https://fal.run/xai/grok-imagine-image

bytedance/seedream/v4.5/...

fal-ai/bytedance/seedream/v4.5/...

https://fal.run/fal-ai/bytedance/seedream/v4.5/...

flux-1/schnell

fal-ai/flux-1/schnell

https://fal.run/fal-ai/flux-1/schnell

Rule:

xai/

models use their ID directly; all others need

fal-ai/

prefix.

2. Get Model Details

Once user confirms a model, fetch its detailed information:

URL Pattern Rules:

Model Prefix	llms.txt URL	Playground URL	Full Model ID
`xai/`	`https://fal.ai/models/xai/<name>/llms.txt`	`https://fal.ai/models/xai/<name>`	`xai/<name>`
Others	`https://fal.ai/models/fal-ai/<id>/llms.txt`	`https://fal.ai/models/fal-ai/<id>`	`fal-ai/<id>`

Rule: If model ID starts with

xai/

, do NOT prepend

fal-ai/

. For all other models, prepend

fal-ai/

.

Examples:

xai/grok-imagine-image

→ llms.txt:

.../models/xai/grok-imagine-image/llms.txt

, Playground:

.../models/xai/grok-imagine-image

bytedance/seedream/v4.5/text-to-image

→ llms.txt:

.../models/fal-ai/bytedance/seedream/v4.5/text-to-image/llms.txt

, Playground:

.../models/fal-ai/bytedance/seedream/v4.5/text-to-image

When presenting model details to user, always include:

Model ID
Playground link: For interactive testing
API endpoint:
```
https://fal.run/<full-model-id>
```
Documentation links: llms.txt, API docs (
```
https://fal.ai/models/<full-model-id>/api
```
)

Process:

Parse the llms.txt content for:
- Model capabilities and use cases
- API endpoint and parameters
- Code examples
- Pricing information
Present the key details to the user
Confirm integration approach with user

3. Integration Guidelines

When integrating a Fal AI model:

Check existing Fal client setup in the project
Use the model ID returned from search (format:
```
fal-ai/<model-name>
```
)
Follow the API patterns shown in the llms.txt examples
Add proper error handling for API calls

4. User Confirmation Points

Always confirm with user at these stages:

After search results: "Found these models: [list with links]. Which one would you like to use?"
After model details: "Here are the details for [model] ([playground link]). Shall I proceed with integration?"
Before code changes: "I'll integrate [model] by [changes]. Does this look correct?"

Link format for presentation:

Use markdown format:
```
[description](url)
```
Always include the model's playground link so users can test it interactively

Common Model Patterns

Model Type	Example IDs	Use Case
Text to Image	`fal-ai/flux-schnell` , `xai/grok-imagine-image`	Generate images from text
Image to Image	`fal-ai/flux-lora` , `xai/grok-imagine-image/edit`	Transform existing images
Video	`fal-ai/ltx-video` , `fal-ai/veo2`	Generate video content
Audio	`fal-ai/playht` , `fal-ai/xtts`	Text-to-speech, audio generation

Example Output Format

When presenting search results to the user, use this format:

Found 3 models matching "flux":

1. **fal-ai/flux-schnell** (Text to Image)
   - Fastest Flux model for high-quality image generation
   - [Playground](https://fal.ai/models/fal-ai/flux-schnell) | [API Docs](https://fal.ai/models/fal-ai/flux-schnell/api)

2. **fal-ai/flux-pro/v1.1** (Text to Image)
   - Professional-grade image generation with superior quality
   - [Playground](https://fal.ai/models/fal-ai/flux-pro/v1.1) | [API Docs](https://fal.ai/models/fal-ai/flux-pro/v1.1/api)

Which model would you like to use?

Error Handling

Error	Cause	Solution
No search results	Query too specific or no matching models	Suggest broader search terms or alternative keywords
Model not found (404)	Wrong URL pattern	Check if model starts with `xai/` → use `/models/xai/...` URL; otherwise use `/models/fal-ai/...`
llms.txt returns empty	Model deprecated or unavailable	Inform user the model may be deprecated; suggest alternatives
API key error (401)	Missing or invalid `FAL_KEY`	Check environment variable or prompt user to set `FAL_KEY`
Rate limit (429)	Too many requests	Implement exponential backoff; suggest retry after delay
Timeout	Model inference taking too long	For async models, use `fal.queue.submit` pattern with polling

Notes

The
```
llms.txt
```
endpoint returns machine-readable documentation optimized for LLMs
Model IDs can have various prefixes:
```
fal-ai/
```
,
```
xai/
```
, etc. - always use the full ID from search results
Always present options to the user before making integration decisions