Trigger Phrase (must be included in the prompt for the full effect)

v13: casual snapshot

v12: casual snapshot

v11: snapshot photo

Recommended inference settings/workflow

use the default "9B Text-to-Image" ComfyUI workflow (found here) with an added "Load LoRa" node and the following settings (which are the default settings for the model as per its HuggingFace page):

euler (sampler), 50 steps, cfg 4, 1 megapixel resolution (1024x1024)

How to prompt:

Ask your LLM of choice to generate you a "literal, technical" prompt that is about 2 paragraphs long of which each paragraph has about 4 sentences that are written in a natural language style.

Latest version notes and model page updates:

major changes: completely revamped entire training workflow, everything is better: quality, coherence, style likeness, etc...

added a "How to prompt" section

LoRa description

This LoRa model was designed to generate images that look like casual snapshot photos taken with a Samsung Galaxy A52 smartphone.