FlowMap is an end-to-end differentiable method that estimates precise camera poses, camera intrinsics, and per-frame dense depth from a video sequence. It minimizes a least-squares objective that compares the optical flow induced by depth, intrinsics, and poses against off-the-shelf optical flow and point track correspondences. The method introduces differentiable re-parameterizations of depth, intrinsics, and pose, enabling first-order optimization. FlowMap outperforms prior gradient-descent-based bundle adjustment methods and performs on par with COLMAP, a state-of-the-art Structure-from-Motion (SfM) method, on the task of 360° novel view synthesis using Gaussian Splatting. FlowMap's outputs enable high-quality photorealistic novel view synthesis, demonstrating its effectiveness in various applications. The method is fully differentiable and can be integrated into deep learning pipelines, making it a promising approach for 3D reconstruction and scene understanding.FlowMap is an end-to-end differentiable method that estimates precise camera poses, camera intrinsics, and per-frame dense depth from a video sequence. It minimizes a least-squares objective that compares the optical flow induced by depth, intrinsics, and poses against off-the-shelf optical flow and point track correspondences. The method introduces differentiable re-parameterizations of depth, intrinsics, and pose, enabling first-order optimization. FlowMap outperforms prior gradient-descent-based bundle adjustment methods and performs on par with COLMAP, a state-of-the-art Structure-from-Motion (SfM) method, on the task of 360° novel view synthesis using Gaussian Splatting. FlowMap's outputs enable high-quality photorealistic novel view synthesis, demonstrating its effectiveness in various applications. The method is fully differentiable and can be integrated into deep learning pipelines, making it a promising approach for 3D reconstruction and scene understanding.