Common Model Arguments

When you call a model on fal, the arguments dict (Python) or input object (JavaScript) becomes the request body. Over HTTP, you send the same fields as the JSON body. Most fields are model inputs. fal also reads some fields as request metadata, such as the user safety identifier. Each model defines its own input schema. You can find that schema on the model’s API page, such as the Nano Banana 2 API. Many image generation models on fal share a set of common arguments that behave consistently across models. These are not platform-level controls (those are platform headers passed via the headers dict). They are model inputs that affect what the model generates and how. This page documents the most common ones for image generation so you don’t have to look them up on every model’s API page individually.

seed

Random seed for reproducible outputs. The same seed with the same inputs produces the same result. Useful for debugging, A/B testing, or letting users “lock” a generation they like and iterate on the prompt.


Type	Integer or null
Default	Random

result = fal_client.subscribe("fal-ai/nano-banana-2", arguments={
    "prompt": "a sunset",
    "seed": 42
})

const result = await fal.subscribe("fal-ai/nano-banana-2", {
  input: { prompt: "a sunset", seed: 42 }
});

num_images

Number of images to generate per request. Higher values produce more results in a single call but increase processing time and cost proportionally.


Type	Integer
Default	1
Range	Typically 1-4

result = fal_client.subscribe("fal-ai/nano-banana-2", arguments={
    "prompt": "a sunset",
    "num_images": 4
})

const result = await fal.subscribe("fal-ai/nano-banana-2", {
  input: { prompt: "a sunset", num_images: 4 }
});

image_size

Output image dimensions, used by models built with fal’s toolkit (such as FLUX models). You can pass a preset string for common aspect ratios, or a custom {width, height} object for precise control.

Preset	Dimensions
`square_hd`	1024x1024
`square`	512x512
`portrait_4_3`	768x1024
`portrait_16_9`	576x1024
`landscape_4_3`	1024x768
`landscape_16_9`	1024x576

result = fal_client.subscribe("fal-ai/flux/schnell", arguments={
    "prompt": "a sunset",
    "image_size": "landscape_16_9"
})

result = fal_client.subscribe("fal-ai/flux/schnell", arguments={
    "prompt": "a sunset",
    "image_size": {"width": 1280, "height": 720}
})

const result = await fal.subscribe("fal-ai/flux/schnell", {
  input: { prompt: "a sunset", image_size: "landscape_16_9" }
});

const result = await fal.subscribe("fal-ai/flux/schnell", {
  input: { prompt: "a sunset", image_size: { width: 1280, height: 720 } }
});

aspect_ratio

Some models (such as Nano Banana 2) use aspect_ratio instead of image_size. This accepts a ratio string and the model determines the output resolution internally.


Type	String
Default	Varies by model
Common values	`"1:1"`, `"16:9"`, `"9:16"`, `"4:3"`, `"3:4"`, `"3:2"`, `"2:3"`, `"21:9"`, `"4:5"`, `"5:4"`

result = fal_client.subscribe("fal-ai/nano-banana-2", arguments={
    "prompt": "a sunset",
    "aspect_ratio": "16:9"
})

const result = await fal.subscribe("fal-ai/nano-banana-2", {
  input: { prompt: "a sunset", aspect_ratio: "16:9" }
});

Check the model’s API page to see whether it uses image_size or aspect_ratio. Passing the wrong one will have no effect.

user (safety identifier)

Provide a stable, privacy-preserving identifier for the end user who initiated the request. fal uses it as safety metadata to connect reported content to that end user. The identifier does not change model behavior or configure safety checks.


Type	String
Required	No
Maximum length	256 characters

Use the same value for the same end user across requests. Hash or pseudonymize the identifier before you send it. Do not include names, email addresses, or other personal information. For users who are not signed in, you can use a stable session identifier. fal includes the identifier in related safety reports. Keep a mapping in your system if you need it to investigate and take action.

result = fal_client.subscribe("fal-ai/nano-banana-2", arguments={
    "prompt": "a sunset",
    "user": "user_7f3a9c"
})

const result = await fal.subscribe("fal-ai/nano-banana-2", {
  input: { prompt: "a sunset", user: "user_7f3a9c" }
});

enable_safety_checker

fal enables safety filtering by default. On models that expose enable_safety_checker, you can use this input to adjust the filter. fal must authorize your account before false can request relaxed filtering. Otherwise, fal keeps filtering on. Contact support to request authorization.


Type	Boolean
Default	`true`

result = fal_client.subscribe("fal-ai/nano-banana-2", arguments={
    "prompt": "a sunset",
    "enable_safety_checker": True
})

const result = await fal.subscribe("fal-ai/nano-banana-2", {
  input: { prompt: "a sunset", enable_safety_checker: true }
});

To request relaxed filtering, set the value to false:

result = fal_client.subscribe("fal-ai/nano-banana-2", arguments={
    "prompt": "a sunset",
    "enable_safety_checker": False
})

const result = await fal.subscribe("fal-ai/nano-banana-2", {
  input: { prompt: "a sunset", enable_safety_checker: false }
});

enable_prompt_expansion

When enabled, fal runs your prompt through an LLM before passing it to the model. The LLM rewrites short or vague prompts into richer descriptions with detail on composition, textures, colors, lighting, and mood. For example, “a sunset” might become a detailed scene description with specific lighting conditions, color palettes, and framing. This typically improves output quality for brief prompts but may alter the result if your prompt is already detailed and precise.


Type	Boolean
Default	Varies by model
Also accepted as	`expand_prompt` (some models use this name instead)
Additional cost	Small per-request charge for the LLM inference step

The expansion adds an LLM inference step before generation, so expect slightly higher latency per request. The expanded prompt is not returned in the response, so if you want to see what the model actually received, run the model with expansion disabled and handle prompt enhancement in your own pipeline.

result = fal_client.subscribe("fal-ai/nano-banana-2", arguments={
    "prompt": "a sunset",
    "expand_prompt": True
})

const result = await fal.subscribe("fal-ai/nano-banana-2", {
  input: { prompt: "a sunset", expand_prompt: true }
});

Check the model’s API page for the exact parameter name. Some models use enable_prompt_expansion, others use expand_prompt.

output_format

Format of the generated image file. Choose based on your use case: jpeg for smaller file sizes with lossy compression, png for lossless quality with transparency support, or webp for a modern format that balances both.


Type	String
Default	Varies by model (typically `"jpeg"` or `"png"`)
Values	`"jpeg"`, `"jpg"`, `"png"`, `"webp"`, `"gif"`

result = fal_client.subscribe("fal-ai/flux/schnell", arguments={
    "prompt": "a sunset",
    "output_format": "png"
})

const result = await fal.subscribe("fal-ai/flux/schnell", {
  input: { prompt: "a sunset", output_format: "png" }
});

Setting Up

Model APIs

Serverless

Compute

Organizations

fal Agent

fal Assets

Common Model Arguments

seed

num_images

image_size

aspect_ratio

user (safety identifier)

enable_safety_checker

enable_prompt_expansion

output_format