Understanding Image Prompting

Stable Diffusion benefits greatly from prompts that use comma-separated words and phrases. This structure helps organize the prompt into clear segments, allowing the model to interpret each element individually. For example, writing “a cat, a dog, a tree” leads to more coherent results than using a long descriptive sentence. This method is especially helpful for users who want to emphasize specific components in their generated images.

In contrast, Flux models perform best when given full sentences that describe the image. Comma-separated tags are not recommended. This marks a key difference between Stable Diffusion and Flux.

LoRA is a fine tuning method that teaches a model to recognize and replicate specific styles, characters, or objects. LoRA models require trigger phrases to activate properly. For example, to generate an accurate result using a logo model, you would need to include a phrase like “Cinna logo.” Some LoRA models may need multiple phrases to trigger effectively. It is recommended that prompts always include the “autofill” field from the model’s configuration and use key terms listed in the “recommend” section.

For best results, we suggest visiting Cinna Imagine and reviewing example prompts on each model's dedicated page.

Need More Models? Image generation models from platforms like Civitai and LLMs from HuggingFace can be added upon request. If you are interested in hosting your own models within Cinna or creating tailored models for your use case, contact us at [e-mail]