Diffusiondb

Visit Tool

DiffusionDB is a large-scale text-to-image prompt dataset based on Stable Diffusion. It contains 14 million images generated by real users, providing a rich resource for research and development.

Claim this tool

No Views Yet

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is diffusiondb?

DiffusionDB is the first large-scale text-to-image prompt dataset, offering 14 million images generated by Stable Diffusion. These images are accompanied by the prompts and hyperparameters specified by real users, making it an invaluable resource for researchers. The dataset is available in two subsets, DiffusionDB 2M (2 million images, 1.6TB) and DiffusionDB Large (14 million images, 6.5TB), catering to different needs. It supports research in understanding prompt-model interplay, deepfake detection, and designing human-AI interaction tools. Users can access the dataset via Hugging Face Datasets, a Python downloader script, or by directly querying the metadata tables for text-only analysis.

Best used for

Ideal for data scientists and developers who need to research the interplay between prompts and generative models, detect deepfakes, and design human-AI interaction tools. Especially valuable for those requiring a massive, human-actuated dataset for advanced machine learning and AI development.

Common actions

analyze image generation

research generative models

detect deepfakes

access image metadata

"AI Agents"github copilotface swappingdeepfakelow-code/no-codeopen-sourceworkflowsautomated workflowcollaboration

Capabilities

Key features

14 million Stable Diffusion images
User-specified prompts
Hyperparameters included
Two dataset subsets
Metadata tables
Hugging Face Datasets integration

Target Audience

data scientistdeveloper

Integrations

hugging-face

Pricing & Plans

Open Source

Free

FAQs

What is the difference between DiffusionDB 2M and DiffusionDB Large?

DiffusionDB 2M contains 2 million images and is 1.6TB, while DiffusionDB Large contains 14 million images and is 6.5TB. DiffusionDB Large is a superset of 2M, and images in 2M are PNG, while Large uses lossless WebP format.

How can I access the DiffusionDB dataset?

You can access DiffusionDB through Hugging Face Datasets, use a provided Python downloader script to download specific parts, or directly query the metadata.parquet files for text-only access to prompts and hyperparameters.

Does DiffusionDB contain NSFW content?

Yes, despite Stable Diffusion's NSFW filter, DiffusionDB may contain some NSFW images. The dataset provides NSFW scores for both images and prompts, allowing users to filter content based on their desired threshold.

Trending

Subcategories trending in Data & Analytics

Business Intelligence Predictive Analytics Real-Time Analytics Market Research Data Cleaning & Prep Data Pipelines & Integration

Trending

Explore

Browse AI tools by category

Content & Design Productivity & Business Coding & Development AI Agents & Automation Research & Education Wellness & Lifestyle Career Development Marketing & Growth Data & Analytics Customer Support & CX Finance E-commerce