How to Access Stable Diffusion 3.5? - Analytics Vidhya

Stable Diffusion 3.5: A Deep Dive into its Models and Access Methods

Stability.ai's latest release, Stable Diffusion 3.5, boasts three customizable models—Large, Large Turbo, and Medium—all optimized for consumer hardware. This article explores these models, their access methods, and their capabilities.

Model Overview

• Availability: Downloadable from Hugging Face and accessible via Stability AI's API, Replicate, and other platforms.
• Safety: Stability AI has implemented safety protocols to mitigate misuse.
• Future Development: Planned enhancements include ControlNet support for finer image control.
• Platform Compatibility: Seamless integration across various platforms ensures flexible usage.

Table of Contents

• Overview
• Stable Diffusion 3.5 Models
• Model Comparison
• Accessing Stable Diffusion 3.5
  • Via Stability.ai Platform
  • Via Hugging Face
• Conclusion
• Frequently Asked Questions

Stable Diffusion 3.5 Models Explained

1. Stable Diffusion 3.5 Large: This flagship model (8.1 billion parameters) delivers superior image quality and prompt adherence, ideal for professional use at 1 megapixel resolution.
2. Stable Diffusion 3.5 Large Turbo: A faster, streamlined version of the Large model. It achieves high-quality results in just 4 steps.
3. Stable Diffusion 3.5 Medium: Designed for consumer hardware (2.5 billion parameters), this model balances quality with efficiency, supporting resolutions from 0.25 to 2 megapixels using the improved MMDiT-X architecture. It requires approximately 9.9 GB VRAM (excluding text encoders).

All models are fine-tunable and optimized for consumer hardware.

Model Comparison

Stable Diffusion 3.5 Large excels in prompt adherence and rivals larger models in image quality. Large Turbo prioritizes speed without sacrificing quality, while Medium provides a high-performing, resource-efficient option.

Accessing Stable Diffusion 3.5

Via Stability.ai Platform

Obtain your API key from the platform (25 credits provided upon signup). Use the following Python code (replace with your API key) within a Jupyter environment:

import requests

response = requests.post(
    f"https://api.stability.ai/v2beta/stable-image/generate/sd3",
    headers={
        "authorization": f"Bearer sk-{API-key}",
        "accept": "image/*"
    },
    files={"none": ''},
    data={
        "prompt": "A middle-aged man wearing formal clothes",
        "output_format": "jpeg",
    },
)

if response.status_code == 200:
    with open("./man.jpeg", 'wb') as file:
        file.write(response.content)
else:
    raise Exception(str(response.json()))

(Example image generated with the prompt: "A middle-aged man wearing formal clothes")

Via Hugging Face

Access the model directly on Hugging Face. The interface allows for immediate inference, as shown below:

(Example image generated with the prompt: "A forest with red trees")

For API access via Hugging Face, follow these steps:

1. Locate the Stable Diffusion 3.5-large model page on Hugging Face. (Note: Other models are selectable).
2. Obtain access (it's a gated model).
3. Use the following Python code (replace with your Hugging Face token):

import requests

API_URL = "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-3.5-large"
headers = {"Authorization": "Bearer hf_token"}

def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.content

image_bytes = query({
    "inputs": "A ninja sitting on top of a tall building, 8k",
})

import io
from PIL import Image
image = Image.open(io.BytesIO(image_bytes))
image

(Example image generated with the prompt: "A ninja sitting on top of a tall building, 8k")

Conclusion

Stable Diffusion 3.5 offers a versatile suite of image generation models catering to various needs and hardware capabilities. Its accessibility through multiple platforms simplifies high-quality AI image generation.

(Consider the GenAI Pinnacle Program for Generative AI training.)

Frequently Asked Questions

Q1: Stability AI API Authentication? Use your API key in the request headers.

Q2: Common Stability AI API Errors? Unauthorized access, invalid parameters, or exceeding usage limits.

Q3: Is Stable Diffusion 3.5 Medium free? Free under the Stability Community License for research, non-commercial use, and organizations with under $1M revenue. Larger entities require an Enterprise License.

Q4: What distinguishes Stable Diffusion 3.5 Medium? Improved MMDiT-X architecture with QK-normalization and dual attention for enhanced image generation across resolutions.

The above is the detailed content of How to Access Stable Diffusion 3.5? - Analytics Vidhya. For more information, please follow other related articles on the PHP Chinese website!