Amateur Selfies is a LoRA trained on ~1000 casual self-portrait photos, designed to capture the look, feel, and imperfections of real-world selfies. It emphasizes natural framing, handheld angles, everyday environments, and the aesthetics of self shot phone photographs.Recommended settings after more experimentation<ul><li>LoRA strength: 0.9 - 1.0. I accidentally generated the 'redux' showcase images with 0.9 and they're fine.</li><li>FP8 or BF16 UNET. GGUF quants cause checkboard artifacts, do not recommend! FP8 can actually 'enhance' the amateur look</li><li>50 steps</li><li>CFG 3.0</li><li>euler/beta</li><li>&lt;= 2.0 MP</li></ul>Workflow is included in the sample images.I have identified a couple of issues with training captions and masks:<ul><li>captions were natural language formatted in paragraphs. The newline paragraph separators however are interpreted by SimpleTuner as delimiter for individual captions.</li><li>masks weren't applied due to a derp in the dataset config</li></ul>I'm working on training v2.0 now using the same image set, but with new captions and fixed masks (face removal). Also appending some quality captions which should improve the sharpness of outputs when not prompting specifically for poor quality / compression artifacts.

Amateur Selfies

Same image dataset used as v1.I fixed the config which improperly specified the masks (face removal only) and improved the captioning to indicate quality of the images: blurriness, compression artifacts etc.The result was a much sharper model which was able to be trained for longer before it collapsed.

Files with SHA256: e1b8b18edcb997297fdb7fe404a1636322ade6138a14040aeb4de7b185cfb64a