Wildfie

Wildfie — Your selfie, a little wilder. Live app: https://wildfie.vercel.app/ Wildfie is a one-click creative portrait experience: upload a selfie or a clear group photo, then see the original beside an animal-inspired version. The transformation carries over recognizable details such as head angle, facial proportions, eye color, glasses, hair, clothing and pose, while keeping the original framing and aspect ratio. A line-art human-to-animal animation makes the wait part of the experience. Both images and their short artwork caption download together as one PNG. My domain is caricature: noticing which visible shapes make a particular portrait distinctive, and translating them into another visual language. Wildfie combines visual interpretation, optional subject segmentation and reference-guided image creation into a complete creative workflow. Group photos retain their shared scene, with each clear person considered separately. The private creator studio, demonstrated in the video, supports three complementary teaching methods: compare three generated animal alternatives and choose or suggest another; rate visible proportions from -5 to +5 and mark standout features; or agree with a finished result or correct its animal choice with a reason. Feedback persists locally and can be exported. Optional memory retrieves relevant examples to support local candidate ranking, with 85% independent visual fit and 15% preference support by default. This is retrieval and scoring, not model fine-tuning. Benchmarks compare frozen predictions and saved labels to evaluate matching changes as the feedback collection grows. Starting from an existing human/animal app scaffold, the hackathon work built the Meta-backed transformation pipeline, group-photo workflow, comparison/export experience, and private teaching, memory and evaluation tools. The frontend is React/Expo on Vercel; the backend is ASP.NET Core on IIS. Animal analogies and caption moods are artistic interpretations, not personality or emotion assessments.