High-fidelity 3D foot reconstruction is crucial for prescription orthotics but is hindered by expensive, specialized equipment that limits patient access. We overcome this barrier with the first end-to-end pipeline to reconstruct clinically-accurate foot meshes from simple, self-captured smartphone videos. Our method uniquely solves the core challenges of in-the-wild scanning: we resolve pose ambiguities using SE(3) canonicalization with view-point prediction, and then complete partial geometry using an attention-based network. Clinical validation demonstrates that our reconstructions achieve state-of-the-art accuracy and meet prescription-readiness standards, preserving the anatomical fidelity essential for medical intervention. By democratizing high-quality foot assessment, our work unlocks new opportunities for accessible telemedicine, preventative diabetic care, and personalized orthotic treatment. We will release our dataset online at: https://bestfootforward.netlify.app.