Stable Diffusion Models
SDXL, SD 3.5, SD 1.5, LoRA, ControlNet, and VAE — then open the generator.
Image Models
Nano Banana
Generate with this model →
Nano Banana 2
hotGenerate with this model →
Nano Banana 2.5
NewGenerate with this model →
Nano Banana 3
newGenerate with this model →
Nano Banana Pro
Generate with this model →
GPT Image 1.5
Generate with this model →
GPT Image 2
newGenerate with this model →
GPT Image 2.5
NewGenerate with this model →
Seedream 4.0
Generate with this model →
Seedream 4.5
Generate with this model →
Seedream 5.0 Lite
Generate with this model →
Seedream 5.0 Pro
hotGenerate with this model →
GLM Image
Generate with this model →
Video Models
Seedance 2.5
NewGenerate with this model →
Seedance 2.0
Generate with this model →
Seedance 1.5 Pro
Generate with this model →
Veo 3.1
Generate with this model →
Sora 2
Generate with this model →
Sora 2 Pro
Generate with this model →
Kling 2.6
Generate with this model →
Kling 3.0
newGenerate with this model →
Kling 3.0 Motion Control
Generate with this model →
Grok Imagine
Generate with this model →
Pixverse V6
Generate with this model →
Pixverse C1
Generate with this model →
Happy Horse 1.0
Generate with this model →
Minimax H3
Generate with this model →
Wan 3.0
Generate with this model →
Stable Diffusion models explained for online creators
Choosing between SDXL, SD 3.5, SD 1.5, LoRA, ControlNet, and VAE decides how well your prompts, composition, and style hold up. This page is a practical Stable Diffusion model guide for creators who generate in the browser on stable diffussion.
Use the cards below to understand each family, then open Generate to run the currently available cloud options. For deep custom checkpoints, pair this guide with Tutorials and Blog deep-dives.
How to pick a model for your task
- Define the goal: portrait, product, anime, landscape, or layout-controlled edit.
- Prefer SDXL for a mature LoRA ecosystem; try SD 3.5 when prompt adherence and text matter more.
- Add one LoRA at moderate weight; stack only after a clean base image.
- Use ControlNet when pose, depth, or edges must match a reference sketch or photo.
Model tips that save credits
- Draft with fewer steps or Turbo-style modes, then finalize with a higher-quality preset.
- Keep aspect ratio consistent when comparing models so results stay fair.
- If faces or hands fail, fix with inpaint instead of endlessly swapping base models.
- Read Blog SDXL vs SD 3.5 and LoRA stacking guides before buying heavy credit packs.
What you learn on this page
- ✓ SDXL / SD 3.5 / SD 1.5 roles in modern Stable Diffusion
- ✓ When to use ControlNet, LoRA, and VAE
- ✓ Links into Generate, Tutorials, and Blog
- ✓ Checkpoint cards deep-link into Generate with matching model ids
Related Stable Diffusion hubs
Continue with generation tools, prompts, and learning paths.
Model FAQ
SDXL vs SD 3.5?
SDXL has the broadest ecosystem; SD 3.5 excels at complex prompts and text. Draft with Turbo-style speed when available, finalize with the higher-quality preset.
What LoRA weight should I use?
Style LoRAs often start at 0.5–0.7; higher values can overfit or clash.
Do I need to download models?
This page is an ecosystem guide. The generate page lists currently available cloud models; for deep customization, run checkpoints locally.
Is every model listed available in the online generator?
Checkpoint cards (SDXL, SD 1.5, SD 2.1, SD 3.5, Turbo) open Generate with that model preselected. ControlNet and LoRA cards are ecosystem guides — open Tutorials to learn how they work.
What is a VAE and when do I care?
A VAE affects color and decode quality. Online presets usually bake a sensible VAE; locally you may swap VAEs for punchier color.
Should beginners start with ControlNet?
Learn txt2img and prompt basics first. Add ControlNet when you need pose or layout lock from a sketch or photo.
stable diffusion is an independent site and is not affiliated with Stability AI.