The year/Independent research

Paper 2510.08673

Thinking with Camera: A Unified Multimodal Model for Camera-Centric Understanding and Generation

Published
Oct 2025
Research lab
Independent
Citations
15
GitHub
420 stars

01 In brief

Summary

The paper introduces Puffin, a unified multimodal model that jointly performs camera-centric understanding and generation.

It treats camera parameters as a first-class modality, introducing 'thinking with camera' to align spatially grounded visual cues with professional photographic terms.

Puffin integrates language regression and diffusion-based generation, trained on Puffin-4M, a dataset of 4 million vision-language-camera triplets.

The model uses both discrete camera tokens and continuous pixel-wise camera maps for flexible spatial control.

Experiments show Puffin outperforms specialized models on camera understanding (e.g., on MegaDepth, TartanAir, LaMAR) and camera-controllable generation (e.g., vs.

PreciseCam, GPT-4o, Qwen-Image).

It also supports cross-view tasks like spatial imagination, world exploration, and photographic guidance via instruction tuning.

The authors release code, models, dataset pipeline, and benchmarks.

Limitations include fixed 512x512 resolution and reliance on offline calibration for evaluation, with future work on multi-scale training and video extension.

02 From the paper

Abstract

Camera-centric understanding and generation are two cornerstones of spatial intelligence, yet they are typically studied in isolation. We present Puffin, a unified camera-centric multimodal model that extends spatial awareness along the camera dimension. Puffin integrates language regression and diffusion-based generation to interpret and create scenes from arbitrary viewpoints. To bridge the modality gap between cameras and vision-language, we introduce a novel paradigm that treats camera as language, enabling thinking with camera. This guides the model to align spatially grounded visual cues with photographic terminology while reasoning across geometric context. Puffin is trained on Puffin-4M, a large-scale dataset of 4 million vision-language-camera triplets. We incorporate both global camera parameters and pixel-wise camera maps, yielding flexible and reliable spatial generation. Experiments demonstrate Puffin superior performance over specialized models for camera-centric generation and understanding. With instruction tuning, Puffin generalizes to diverse cross-view tasks such as spatial imagination, world exploration, and photography guidance. We will release the code, models, dataset pipeline, and benchmark to advance multimodal spatial intelligence research.