Checkpoint
Image & Video AIA saved snapshot of a trained AI model's weights at a specific point, commonly used in Stable Diffusion to produce different art styles.
In AI image generation, a checkpoint is a saved version of a model's learned knowledge. Think of it as a 'save file' for the AI's brain. Different checkpoints produce different styles — one might excel at photorealism, another at anime, another at oil paintings.
Checkpoints are central to the Stable Diffusion ecosystem. Community platforms like Civitai host thousands of custom checkpoints that users have fine-tuned for specific aesthetics. You download a checkpoint, load it into your Stable Diffusion interface (like Fooocus or ComfyUI), and it changes what the model 'knows' about art.
Checkpoints are large files (2-7 GB each) because they contain the full model weights. They differ from LoRAs, which are smaller add-ons that modify a base checkpoint rather than replacing it entirely.
Real-World Example
On Civitai you can browse thousands of Stable Diffusion checkpoints — each producing a distinct visual style from photorealistic to anime.
Related Terms
More in Image & Video AI
FAQ
What is Checkpoint?
A saved snapshot of a trained AI model's weights at a specific point, commonly used in Stable Diffusion to produce different art styles.
How is Checkpoint used in practice?
On Civitai you can browse thousands of Stable Diffusion checkpoints — each producing a distinct visual style from photorealistic to anime.
What concepts are related to Checkpoint?
Key related concepts include Stable Diffusion, LoRA (Low-Rank Adaptation), Fine-tuning. Understanding these together gives a more complete picture of how Checkpoint fits into the AI landscape.