Guian Fang

I am a Ph.D. student at Show Lab, National University of Singapore, advised by Prof. Mike Zheng Shou.

Previously, I received my B.Eng. in HCPLab, Artificial Intelligence at the School of Intelligent Systems Engineering, Sun Yat-sen University, advised by Xiaodan Liang (梁小丹), co-supervised by Shengcai Liao.

I'm currently building agentic visual generation pipelines and runtime infrastructure for coding agents. Open to collabs on long-form video, world models, embodied intelligence, and agent orchestration — feel free to reach out.

News

2026.05 Open-sourced AnyFlow — any-step video diffusion via flow map distillation.
2026.05 Released Claw Orchestrator — a unified runtime for Claude Code, Codex and other coding CLIs.
2026.03 Launched PAI at Utopai Studios — long-form video generation for cinematic storytelling.
2025.07 RealignDiff accepted to IEEE TNNLS — coarse-to-fine semantic re-alignment for diffusion.
2025.06 Launched MikoAI — an ACG (anime / comic / manga) creation tool, powered by FramePrompt.
2024.08 Started Ph.D. at Show Lab, NUS, advised by Prof. Mike Zheng Shou.
2024.07 HumanRefiner accepted to ECCV 2024.
2024.06 Co-organized LOVEU workshop at CVPR 2024 — long-form video understanding & AI copilots.
2024.05 ChartThinker presented at LREC-COLING 2024 in Torino.
2023.07 Released LLaMA2-Accessory with OpenGVLab, Shanghai AI Lab — open-source LLM toolkit.

Research & Projects

My research focuses on Generative Models for Vision — particularly long-form video generation, video diffusion, and Video World Models for video understanding and generation — as well as Embodied Intelligence. Recently I've been exploring agentic visual generation, building creative pipelines from concept to final frame, and runtime infrastructure for coding agents. Representative works are highlighted. * denotes equal contribution.

AnyFlow:
Any-Step Video Diffusion Model with On-Policy Flow Map Distillation

Yuchao Gu, Guian Fang, Yuxin Jiang, Weijia Mao, Song Han, Han Cai, Mike Zheng Shou

arXiv preprint, 2026

Paper Project page Models

The first any-step video diffusion framework built on flow maps. AnyFlow shifts distillation from endpoint consistency mapping to flow-map transition learning over arbitrary intervals, with Flow Map Backward Simulation for efficient on-policy distillation. One model adapts to arbitrary step budgets across causal and bidirectional architectures, scaling from 1.3B to 14B parameters on T2V, I2V, and V2V.

Claw Orchestrator banner — unified runtime for coding agent CLIs

Claw Orchestrator:
A Unified Runtime for Coding Agent CLIs

Guian Fang

Open-source, 2026

A TypeScript runtime that turns coding CLIs (Claude Code, Codex, Gemini, Cursor Agent, OpenCode, and custom engines) into persistent, programmable agents — with sessions, multi-engine routing, and multi-agent councils behind one API. Runs standalone, with first-class OpenClaw plugin support.

PAI: Long-form Video Generation for Cinematic Storytelling

Utopai Studios*, Guian Fang*

Productized at Utopai Studios, 2026

Launch post Product page

A productized long-form video generation system for minute-level cinematic continuity. PAI maintains characters, tone, and narrative coherence across shots, with creator control from concept to final frame.

ACG Creation Made Easy for Everyone.

MikoAI*, Guian Fang*

Productized at MikoAI, 2025

Technical Report MikoAI

FramePrompt is the character-animation engine behind MikoAI. By packing reference images, skeleton motion, and target clips into one visual sequence, it turns animation into conditional future prediction — driving pre-trained video diffusion transformers with no guider modules or multi-stage pipelines.

Generate 3D Worlds in Production with AI

Cybever*, Guian Fang*

Productized at Cybever, 2024

An image-to-3D pipeline for 3D world generation in production. Cybever handles the heavy lifting — geometry, layout, materials — so 3D professionals can focus on storytelling.

HumanRefiner research visualization showing human pose refinement

HumanRefiner:
Benchmarking Abnormal Human Generation and Refining with Coarse-to-fine Pose-Reversible Guidance

Guian Fang*, Wenbiao Yan*, Yuanfan Guo*, Jianhua Han, Zutao Jiang, Hang Xu, Shengcai Liao, Xiaodan Liang

ECCV, 2024

Dataset page Paper

A new large-scale benchmark (AbHuman) for anatomical anomalies in humans, and a plug-and-play method (HumanRefiner) for refining abnormal human generations with pose-reversible guidance.

ChartThinker framework diagram showing contextual chain-of-thought approach

ChartThinker:
A Contextual Chain-of-Thought Approach to Optimized Chart Summarization

Mengsha Liu, Daoyuan Chen, Yaliang Li, Guian Fang, Ying Shen

LREC-Coling, 2024

Dataset page Paper

ChartThinker leverages chain-of-thought reasoning and context retrieval to generate accurate and coherent chart summaries, outperforming previous methods on a diverse benchmark.

LLaMA2-Accessory toolkit architecture diagram

LLaMA2-Accessory:
An Open-source Toolkit for LLM Development

Chris Liu, Ziyi Lin, Guian Fang, Jiaming Han, Yijiang Liu, Renrui Zhang, Longtian Qiu, Yichi Zhang, Siyuan Huang

Open-source, 2023

Checkpoint page Documentation

An open-source toolkit supporting pretraining, finetuning, and deployment of large language and multimodal models, making LLM development more accessible and efficient.

RealignDiff framework showing coarse-to-fine semantic re-alignment process

RealignDiff:
Boosting Text-to-Image Diffusion Model with Coarse-to-fine Semantic Re-alignment

Guian Fang*, Zutao Jiang*, Jianhua Han, Guansong Lu, Hang Xu, Shengcai Liao, Xiaojun Chang, Xiaodan Liang

IEEE TNNLS, 2023

Paper

RealignDiff introduces a two-stage semantic re-alignment strategy to significantly improve the consistency between generated images and text prompts in diffusion models.

Honors & Awards

Scholarships

National Scholarship (2023)
SYSU Outstanding Student Scholarship — 1st Class (2023)
Li Xuerou Foundation Scholarship (2023)
Huawei Intelligent Foundation Scholarship (2022)
National Encouragement Scholarship (2022)

Competitions

1st Place, National College Computer Ability Challenge (2023)
2nd Place, Asia and Pacific Mathematical Contest in Modeling (2022)
Silver Medal, China Collegiate Algorithm Design & Programming Challenge Contest (2022)
2nd Place, Social Computing Innovation Competition (2022)

Activities & Services

Conference Reviewer

CV: ECCV, CVPR, ICCV
ML: NeurIPS, ICLR, ICML
AI: AAAI, AISTATS
NLP: ACL Rolling Review (ACL, EMNLP, NAACL, EACL)

Workshop Organizer

LOVEU Workshop @ CVPR 2024: Long-form Video Understanding Towards Multimodal AI Assistant and Copilot

Teaching Assistant

EE3703: Machine Learning with Applications
EE4309: Robot Perception
EE5106: Advanced Robotics

Acknowledgements

Grateful to the mentors and teams I've worked with along the way.