ShutterMuse: Capture-Time Photography Guidance with MLLMs

Abstract

Real-world photography requires capture-time guidance for both camera framing and subject pose. Yet existing aesthetic cropping benchmarks mainly evaluate post-hoc crop prediction and overlook subject-side recommendations, leaving the capture-time guidance capabilities of multimodal large language models (MLLMs) underexplored. To address this gap, we introduce CaptureGuide-Bench, a benchmark with two complementary tasks: photographer-side composition decision and refinement, and subject-side scene-conditioned pose recommendation. Our evaluation reveals complementary limitations: general-purpose MLLMs can make composition decisions but lack precise refinement localization, while specialized aesthetic cropping models localize crops effectively but are limited to refinement; neither provides structured, actionable pose guidance. To support model development, we further construct CaptureGuide-Dataset, comprising 130K samples with textual rationales and structured visual annotations, and develop ShutterMuse, a unified MLLM trained with supervised and reinforcement fine-tuning. Experiments on CaptureGuide-Bench show that ShutterMuse achieves state-of-the-art photographer-side guidance and competitive subject-side pose recommendation with substantially lower inference cost, demonstrating the potential of MLLMs as interactive assistants for photography during image capture.

Demo Video

Discover ShutterMuse in action through a comprehensive demonstration of pose recommendation, composition decision-making, and intention-guided cropping.

CaptureGuide Dataset and Bench

CaptureGuide contains two task sides: photographer-side composition guidance and subject-side pose guidance. CaptureGuide-Dataset supports model development, while CaptureGuide-Bench evaluates composition decision/refinement and pose recommendation quality.

Pose Recommendation

Keypoint skeletons generated by ShutterMuse are used to guide GPT-Image-2 in rendering human pose images.

Composition Recommendation

Original photos are paired with ShutterMuse's recommended crops.

Comparison with State-of-the-Art Methods

Qualitative comparison of photographer-side guidance with state-of-the-art methods

Qualitative comparison of subject-side pose guidance with state-of-the-art methods

Citation

@misc{li2026shuttermuse,
  title        = {ShutterMuse: Capture-Time Photography Guidance with MLLMs},
  author       = {Li, Jiayu and Fang, Yixiao and Hu, Tianyu and Cheng, Wei and Huang, Ping and Fan, Zheheng and Yu, Gang and Ma, Xingjun},
  year         = {2026},
  note         = {Preprint}
}