This model was trained on my personal photographs of Hyla cinerea (Green Treefrog), which were intended for use in a field guide project, so generated images may come out having those qualities (or with five legs, or missing digits, sometimes, too! HA!).Base model imagine generators tend to have limited—often cartoonish—representations of amphibians and reptiles...so I decided to train a LoRA that could produce more photographic and accurate results of a favorite local species, although it also renders in a variety of other styles if prompted for or by applying additional style LoRAs. I'm not sure who needs to render AI Green Treefrogs, but now you can! HA! If folks like it, I might train some other LoRA models of amphibians and reptiles from my photo libraries.I don't think the trigger word alone really works well (or much at all), so include "hyla cinerea" and/or "green treefrog" in your prompt. If the frog is not the central and largest character, you will need to apply inpainting to produce a decent result. There aren't any jumping examples in the dataset, so don't expect action shots to render very well. In interactions with humans and other things, it doesn't always render the frog to the proper scale, so you might try to guide it with descriptors (e.g., "small" or "smaller than the human's hands", etc.).<h2 id="recommended-settings:-luibpo1de">Recommended Settings:</h2><ul><li>Checkpoint: <a target="_blank" rel="ugc" href="http://Flux1.Dev">Flux1.Dev</a> (or variants)</li><li>LoRA strength: 1</li><li>Sampling &amp; scheduler: [Forge] Flux Realistic + SGM Uniform</li><li>Steps: 20+ (or 10+ with turbo LoRA)</li><li>Distilled CFG: 2–4</li></ul>

Files with SHA256: 492e2cc1d7c9a6fb05aa131d7a9a18294548a71dbb3af3169f2d1676ab111c99