Image Generation¶

Tarash Gateway provides a unified interface for generating images across multiple providers. Pass a ImageGenerationConfig and an ImageGenerationRequest to generate_image() (or generate_image_async() for async workflows); the response is always a normalized ImageGenerationResponse regardless of which provider handled the request.

Text to Image¶

from tarash.tarash_gateway import generate_image
from tarash.tarash_gateway.models import ImageGenerationConfig, ImageGenerationRequest

config = ImageGenerationConfig(
    provider="fal",
    model="fal-ai/flux/dev",
    api_key="YOUR_FAL_KEY",
)
request = ImageGenerationRequest(
    prompt="A serene mountain lake at sunrise, photorealistic",
)
response = generate_image(config, request)
print(response.images[0])

Fal.ai image models →

Image to Image / Edit¶

Providers that support image-to-image workflows accept an input image via the image_list field on the request. Each entry is a dict with an image key (URL, base64 string, or raw bytes) and a type key describing its role (e.g. "reference", "asset").

from tarash.tarash_gateway import generate_image
from tarash.tarash_gateway.models import ImageGenerationConfig, ImageGenerationRequest

config = ImageGenerationConfig(
    provider="fal",
    model="fal-ai/flux-2-pro",
    api_key="YOUR_FAL_KEY",
)
request = ImageGenerationRequest(
    prompt="Make it look like a watercolor painting",
    image_list=[
        {"image": "https://example.com/input.jpg", "type": "reference"},
    ],
)
response = generate_image(config, request)
print(response.images[0])

Fal.ai image models →