Japanese Stable Diffusion XL
accounts/fireworks/models/japanese-stable-diffusion-xlStability AI
Image
Serverless
Japanese Stable Diffusion XL (JSDXL) is a Japanese-specific SDXL model that is capable of inputting prompts in Japanese and generating Japanese-style images.
Features
Serverless API
DocsJapanese Stable Diffusion XL is available via Fireworks' serverless image API, where you pay per image step (per generated image) and do not need to configure GPUs.
Available Serverless
Run queries immediately, pay only for usage
$0.00013
Per Step# pip install 'fireworks-ai' import fireworks.client from fireworks.client.image import ImageInference, Answer # Initialize the ImageInference client fireworks.client.api_key = "$API_KEY" inference_client = ImageInference(model="accounts/fireworks/models/japanese-stable-diffusion-xl") # Generate an image using the text_to_image method answer : Answer = inference_client.text_to_image( prompt="A beautiful sunset over the ocean", cfg_scale=undefined, height=1024, width=1024, sampler=None, steps=undefined, seed=0, safety_check=False, output_image_format="JPG", # Add additional parameters here ) if answer.image is None: raise RuntimeError(f"No return image, {answer.finish_reason}") else: answer.image.save("output.jpg")
Image API Reference
Calling the API
The Fireworks image generation API can be called by making HTTP requests from any language. We also provide a Python client library that you can use to easily call the API.
Instructions for deploying on-demand deployments for these models can be found on the model HuggingFace pages.
Below we describe the raw REST API for making requests to the API. For easy generation of calls, see the console above.
Authentication
All API calls require a valid authentication token. You can obtain a token from the Fireworks console. When calling the API, make sure to specify the Authentication header, with your token after the word Bearer. For example: Authorization: Bearer abcd1234defg5678hijk9012lmno3456pqrs7890tuvw1234.
File Uploads (image_to_image and control_net)
Since image_to_image and control_net require uploading binary image files, the APIs are designed around the multipart/form-data content type. This means that forming your query is not as simple as writing a JSON response body, as with text-to-image. See information about how to make these requests in Python requests, node.js fetch, and cURL.
A common pitfall with multipart/form-data is to try to explicitly set Content-Type: multipart/form-data in your request. This is incorrect, as the `Content-Type` requires a boundary string to be specified as well. Instead, use the form submission API of the request library you're using, for example, the Python requests modules forms a multipart/form-data request when you include files as a request parameter.
Another common pitfall with multipart/form-data is to erroneously set Content-Type: application/json in your request, especially if you are directly porting from the text-to-image API. The service will return an HTTP 415 Unsupported Media Type response if you specify the wrong `Content-Type`.
Image Generation Modes
Generate an image from a text prompt
POST https://api.fireworks.ai/inference/v1/image_generation/accounts/fireworks/models/japanese-stable-diffusion-xl HTTP/1.1
Request Headers
Request Parameters
Response Codes
Metadata
Ready