Stable Diffusion Models Hub
Stable Diffusion is a family of models, not one: SD 1.5, SDXL, SD 3 and SD 3.5 differ in size, native resolution and license. Two of them run here - SDXL 1.0 and DreamShaper 8, an SD 1.5 fine-tune - both text to image and on every plan, beside the other image and video models below.
Which Stable Diffusion model is best for what?
Three generations of Stable Diffusion models matter when you pick one today. Each entry gives the release, what changed and whether it runs here; the explainer at /blog/what-is-stable-diffusion lists the sources.
Stable Diffusion 1.5 and its fine-tunes
Released in October 2022 and trained at 512×512. Small and fast, with the largest library of community checkpoints. DreamShaper 8 is one of them, and it runs here in its LCM (latent consistency model) build.
SDXL 1.0
Released in July 2023: a UNet about three times larger, a second text encoder and native 1024×1024 output, from a 3.5B-parameter base model. It is the default model here, text to image only, at 4 credits an image.
Stable Diffusion 3 and 3.5
SD 3 Medium (June 2024, 2B parameters) replaced the U-Net with a Diffusion Transformer. SD 3.5 (October 2024) came as Large (8.1B parameters), Large Turbo (a distilled four-step version of Large) and Medium (2.5B). Neither is offered on this site.
Model pages on this site
One page per model family, for images and for video. A model page is a description, not a separate generator: every one of them runs from the same two forms.
Stable Diffusion: SDXL and DreamShaper 8
The two Stable Diffusion models here, both on every plan and both text to image only. /sdxl and /dreamshaper list their size, steps, guidance and seed controls.
Other image models
Flux 2, Flux Kontext, Nano Banana, Nano Banana 2, Nano Banana Pro, Midjourney, GPT Image 1.5, GPT Image 2, GPT Image 2.5, Seedream 4.5, Seedream 5 Lite and Z-Image each have a page. None of them is a Stable Diffusion model.
Video models
Veo 3, Kling, Wan, Runway, Hailuo, Grok Imagine, Midjourney Video, Seedance and Seedance 2.0 Mini. Video runs from /ai-video-generator, not the image form.
What a model page does not do
It does not benchmark models against each other or promise a quality ranking. The comparison a model page supports is inputs, controls, plan and credit cost.
How to choose a Stable Diffusion model
Decide still or motion first
Images and video run from different forms. A still-image model cannot be selected in the video generator and the reverse is also true.
Start with Stable Diffusion XL for images
It is on every plan, costs 4 credits an image and takes a text prompt with an optional negative prompt. Pick DreamShaper 8 at the same price for the SD 1.5 look, and a reference-image model when a photo has to shape the result.
Check the plan before you commit to a model
Most image models outside Stable Diffusion, and most video models, need a paid plan. /ai-models lists every selectable model with its current availability.
Test the prompt, then switch models
The same prompt behaves differently per model. Generate once, then change only the model to see what the model itself is contributing.
Models Hub FAQ
Pick a model, then generate
A model page should answer the input, control, plan and cost question in one read, then send you back to the form.
