<h2 id="v7.2-(2025-06-01)-8wmt7fjsd">v7.2 (2025-06-01)</h2>Quality is improved greatly and it is more realistic. However it can be a little wild and unstable.v7.2 is trained with 50 new images carefully picked, that is closer to real animal mouth anatomy. It contains big cats and canines mainly, but due to different training parameters, it is more flexible than older versions and is suitable for many carnivorous mammals, birds, orca and shark. Extremely closeup is possible. also palate ridges and feline tongue texture is better.Suggested settings<ul><li>Base checkpoint: <a rel="ugc" href="https://civarchive.com/models/941929?modelVersionId=1319871">Flux1-DedistilledMixTuned(v3.0)</a></li><li>Other loras are not necessary, but can be used</li><li>Prompt: "mawshot" is the trigger word. it should also include description regarding to mouth. you can start with "close up of &lt;species&gt;'s head/mouth". you can also see the example images.</li><li>Weight: 0.8. high value may slightly influence style and be less stable</li><li>Flux Guidance: ~2.7 for realistic photo, ~3.5 for anime painting, flat coloring</li><li>a simple workflow is included in the example images</li></ul><h2 id="-dguoco4n5"></h2><h2 id="v6.1-(2025-04-30)-udnrz8id3">v6.1 (2025-04-30)</h2>Trained another new version with 75 images, with:<ul><li>more diverse species including penguin, fox, orca/shark</li><li>more styles like 3d rendering, watercolor painting, comic book art</li><li>object in mouth, mouth holding object, dripping saliva</li></ul>It was trained on fluxmania as the base model, 3600 steps. This version does not use a trigger word, and may be better or worse over v5.1 in different aspects.Suggested settings:<ul><li>No trigger word. Describe the animal's mouth features in your prompt like "wide open mouth", "close up of the mouth", "pink tongue", "sharp teeth and fangs", "saliva dripping", "gaping throat"</li><li>Base Checkpoint: <a target="_blank" rel="ugc" href="https://civarchive.com/models/778691?modelVersionId=1539776">Fluxmania V</a></li><li>LoRa: <a target="_blank" rel="ugc" href="https://civarchive.com/models/310964?modelVersionId=939115">Furry Enhancer</a>, 0.15~0.3 to fix palate ridges</li><li>Weight: 0.5~1.0. It's still good at 1.0</li><li>Flux Guidance: 2~3 for realistic output, ~3.5 for anime painting, flat coloring</li><li>a simple workflow is included in the example images </li></ul><h2 id="v5.1-(2025-04-20)-tgs9g3did">v5.1 (2025-04-20)</h2>This version focuses on more poses/viewing angles, and more realistic, less "burned" result. The teeth shape is also better over the previous version.trained on a new dataset with more diverse poses, 25 images in total. And I trained it with prodigy optimizer and picked the best epoch. This version also can affect general style and fur. It's not as strong as v3.0 and result is good when strength=1.0. I suggest it to be used with Furry Enhancer <a target="_blank" rel="ugc" href="https://civarchive.com/models/310964?modelVersionId=939115">https://civarchive.com/models/310964?modelVersionId=939115</a> with a low weight around 0.3 to fix palate ridges. Flux Guidance can be set lower to 2~3 to get a more realistic result. The example image contains a simple ComfyUI workflow. <h2 id="v3.0-9nycoaof1">v3.0</h2> This is a LoRa model for use with Flux-dev that can enhance feral animal's open mouth. It is suitable for realistic style images. It is good for big cats, bears, canines, but can also do some dragons/reptiles. You should write "mawshot" in the prompt, like "a close up mawshot of a polar bear, (describe your image)". The model is trained with 30 images of a variety of animals. The model might affect style, so decrease the strength (0.4 for example) if you need a specific style. For normal use you can set it to 0.6~0.8. This may sometimes create multiple rows of teeth if set too high. I am new to LoRa training and didn't test it out thoroughly so feel free to experiment more.

Realistic Open Mouth for Feral Animals

this version over v3.0 focuses on more poses/viewing angles, and more realistic, less "burned" result. The teeth shape is also better over the previous version.v5.1 is trained on a new dataset with more diverse poses, 25 images in total. And I trained it with prodigy optimizer and picked the best epoch. This version also can affect general style and fur. It's not as strong as v3.0 and result is good when strength=1.0. I suggest it to be used with Furry Enhancer <a target="_blank" rel="ugc" href="https://civitai.com/models/310964?modelVersionId=939115">https://civitai.com/models/310964?modelVersionId=939115</a> with a low weight around 0.3 to fix palate ridges. Flux Guidance can be set lower to 2~3 to get a more realistic result. The example image contains a simple ComfyUI workflow.

Files with SHA256: 0a888ce9878e54f7566f3f1b9518dd0a5f9e90f847e01bd087ea37f520614977