arxiv: 2604.09195 · v1 · submitted 2026-04-10 · 💻 cs.AI

Recognition: unknown

Camera Artist: A Multi-Agent Framework for Cinematic Language Storytelling Video Generation

Haobo Hu , Qi Mao , Yuanhang Li , Libiao Jin

Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:47 UTC · model grok-4.3

classification 💻 cs.AI

keywords multi-agent frameworkvideo generationcinematic languagestorytelling videonarrative consistencycinematography agentstoryboard generation

0 comments

The pith

A dedicated Cinematography Shot Agent with recursive storyboarding and cinematic language injection improves narrative consistency and film quality in multi-agent video generation.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces Camera Artist as a multi-agent framework that follows real-world filmmaking steps to turn scripts into narrative videos. Existing systems often produce disconnected shots and lack deliberate cinematic choices, so the new approach adds a Cinematography Shot Agent that builds storyboards recursively to link adjacent shots and inserts cinematic language to guide expressive camera work and editing. The authors argue this structured injection of film technique creates more continuous storytelling and higher-quality results than prior pipelines. They support the claim with quantitative metrics and human evaluations showing gains in consistency, expressiveness, and overall film perception.

Core claim

Camera Artist builds upon established agentic pipelines and introduces a dedicated Cinematography Shot Agent, which integrates recursive storyboard generation to strengthen shot-to-shot narrative continuity and cinematic language injection to produce more expressive, film-oriented shot designs.

What carries the argument

The Cinematography Shot Agent, which performs recursive storyboard generation for narrative continuity between shots and injects cinematic language to create more deliberate, expressive shot designs.

If this is right

Generated videos maintain stronger narrative progression across consecutive shots due to the recursive storyboard process.
Individual shots incorporate more deliberate cinematic techniques, increasing dynamic expressiveness.
Human viewers rate the output higher in perceived film quality compared with baseline multi-agent outputs.
The framework demonstrates consistent outperformance on both automated metrics and qualitative assessments of storytelling coherence.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Embedding explicit cinematography rules as structured agent behavior could transfer to other generative media tasks that benefit from domain conventions, such as animation or interactive storytelling.
The recursive storyboard mechanism suggests a general pattern for maintaining long-range coherence in sequential generation systems without requiring full re-planning at every step.
If the cinematic language injection proves robust, similar lightweight domain-knowledge modules might reduce reliance on ever-larger base models for creative control.

Load-bearing premise

The specific combination of recursive storyboard generation and cinematic language injection inside the Cinematography Shot Agent is required to achieve the reported gains in continuity and quality, rather than these gains being obtainable through simpler additions to existing multi-agent systems.

What would settle it

An ablation or extension experiment on a baseline multi-agent pipeline that adds general storyboard planning and shot selection but omits the recursive continuity loop and explicit cinematic language rules, then measures whether narrative consistency and film-quality scores reach the same level as Camera Artist.

Figures

Figures reproduced from arXiv: 2604.09195 by Haobo Hu, Libiao Jin, Qi Mao, Yuanhang Li.

**Figure 1.** Figure 1: Comparison with multi-agent system on filmic storytelling. Existing multi-agent methods tend to exhibit fragmented narratives and weak cinematic control. In contrast, Camera Artist achieves stronger shot-to-shot coherence and richer cinematic expression, yielding more filmic storytelling. MAS frameworks primarily focus on the logical alignment between scripts and visuals, often resulting in a mechanical as… view at source ↗

**Figure 2.** Figure 2: The overall framework of Camera Artist. Camera Artist operates in two stages: footage construction and shot generation. In the footage construction stage, the Director Agent expands the story outline and builds hierarchical storyboard assets at script, scene, and shot levels. In the shot generation stage, the Cinematography Shot Agent first performs recursive shot generation to ensure narrative coherence, … view at source ↗

**Figure 3.** Figure 3: Mechanism of the Cinematography Shot Agent. (a) Recursive Shots Generation (RSG): By recursively generating shots and selecting start/mid/end types, the system produces storyboards with strong narrative coherence. (b) Cinematic Language Injection (CLI): A fine-tuned LLM trained on professional cinematic language transforms original shot descriptions into film-style, cinematically expressive ones. scene-le… view at source ↗

**Figure 4.** Figure 4: Qualitative experimental results of single shot content. For videos with similar shot content, Camera Artist can achieve richer and more expressive cinematic language, outperforming prior multi-agent methods. where, fLLM denotes the LLM mapping function. The optimization objective for LLM fine-tuning is formulated as follows: Lcine = − X N i=1 log Pθ ′ (yi | xi). (3) During inference, the fine-tuned LLM i… view at source ↗

**Figure 6.** Figure 6: User study comparison on four subjective metrics. Results of VGoT [10], Anim-Director [8], MovieAgent [11], and our method on Script Consistency, Camera-Movement Consistency, Video Quality, and Real-Movie Similarity. Our method achieves the highest scores across all metrics. inconsistency: VGoT [10] abruptly shifts from “a forest” to “a lakeside” and MovieAgent [11] transitions from “a nighttime forest” to… view at source ↗

**Figure 7.** Figure 7: Ablation study on RSG and CLI. RSG preserves coherent shot-to-shot narrative flow, while CLI enhances cinematic expressiveness through deliberate camera motion and lighting; removing either results in fragmented storytelling or visually static shots. D. Ablation Study To evaluate the contribution of the core modules, we conduct an ablation study on (i) RSG and (ii) CLI. Quantitative and qualitative resul… view at source ↗

**Figure 8.** Figure 8: Camera Artist workflow visualization. Given a user-provided story outline, Camera Artist decomposes the narrative into structured scene plots and character assets via the Director Agent, refines them into coherent shot-level descriptions with explicit cinematic language using the Cinematography Shot Agent, and finally renders corresponding visual clips through the Video Generation Agent. The collaboration … view at source ↗

**Figure 9.** Figure 9: An example of pipeline for cinematic language LoRA fine-tuning. Ordinary captions are produced by a VLM from raw video, while ShotBench [16] provides shot-level cinematic annotations. A LoRA-tuned LLM learns to transform ordinary captions into cinematic shot descriptions with explicit cinematic language, which are later used for cinematic language injection during inference. is not limited to cases where c… view at source ↗

**Figure 10.** Figure 10: The CoT Description of Camera Artist. (a) The CoT of Director Agent, which is mainly responsible for the expansion of script content and scene splitting.(b) The CoT of Cinematography Shot Agent, which is mainly responsible for the recursive generation of storyboard content and the introduction of shot language [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗

**Figure 11.** Figure 11: The CoT prompting of VLM-based evaluation. The CoT is mainly responsible for the recursive generation of storyboard content and the introduction of shot language [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗

**Figure 12.** Figure 12: Qualitative comparison with baseline methoda. (a) Camera Artist generates a final wide shot with high-angle composition and slow pull-back movement, delivering stronger cinematic atmosphere and expressive visual storytelling. (b) Baselines introduce irrelevant characters or exhibit abrupt narrative jumps in two-shot sequences, while Camera Artist maintains both character/scene consistency and coherent eve… view at source ↗

**Figure 13.** Figure 13: Reference-free storytelling video generation. Given only a textual story outline (no character reference images), Camera Artist automatically constructs scenes, characters, and shot sequences, producing a long-form narrative video with coherent story progression and cinematic visual expression [PITH_FULL_IMAGE:figures/full_fig_p012_13.png] view at source ↗

**Figure 14.** Figure 14: Additional qualitative results. Scene-level keyframes together with the corresponding footage are presented, illustrating coherent long-range storytelling, consistent character depiction, and film-style visual expression [PITH_FULL_IMAGE:figures/full_fig_p013_14.png] view at source ↗

read the original abstract

We propose Camera Artist, a multi-agent framework that models a real-world filmmaking workflow to generate narrative videos with explicit cinematic language. While recent multi-agent systems have made substantial progress in automating filmmaking workflows from scripts to videos, they often lack explicit mechanisms to structure narrative progression across adjacent shots and deliberate use of cinematic language, resulting in fragmented storytelling and limited filmic quality. To address this, Camera Artist builds upon established agentic pipelines and introduces a dedicated Cinematography Shot Agent, which integrates recursive storyboard generation to strengthen shot-to-shot narrative continuity and cinematic language injection to produce more expressive, film-oriented shot designs. Extensive quantitative and qualitative results demonstrate that our approach consistently outperforms existing baselines in narrative consistency, dynamic expressiveness, and perceived film quality.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Camera Artist adds a targeted Cinematography Shot Agent with recursive storyboarding and cinematic language injection, but the outperformance claims rest on unablated comparisons that leave the real source of gains unclear.

read the letter

The core contribution is a multi-agent setup that adds one focused agent for shot-level cinematography. It uses recursive storyboard generation to link adjacent shots and injects explicit film terms like angles and lighting to push outputs toward more deliberate cinematic style. That combination is not just another generic agent pipeline; it directly targets continuity and expressiveness gaps that standard script-to-video flows still show.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes Camera Artist, a multi-agent framework for generating narrative storytelling videos with explicit cinematic language. It extends prior agentic pipelines by introducing a dedicated Cinematography Shot Agent that performs recursive storyboard generation to improve shot-to-shot narrative continuity and cinematic language injection to enhance dynamic expressiveness and filmic quality. The authors state that extensive quantitative and qualitative results show consistent outperformance over existing baselines in narrative consistency, dynamic expressiveness, and perceived film quality.

Significance. If the empirical claims hold under controlled evaluation, the work could advance multi-agent systems for creative video generation by providing structured mechanisms for cinematic techniques and narrative progression, addressing fragmentation issues in current approaches to automated filmmaking workflows.

major comments (2)

Abstract: The claim that the approach 'consistently outperforms existing baselines' in narrative consistency, dynamic expressiveness, and perceived film quality is not accompanied by any specific metrics, baseline names, dataset details, or ablation results. This makes it impossible to verify whether gains are attributable to the Cinematography Shot Agent's recursive storyboard generation and cinematic language injection rather than confounding factors such as base model choice or overall prompt complexity.
Experiments section (inferred from abstract's reference to quantitative/qualitative results): No ablation studies are described that isolate the recursive storyboard generation (e.g., full model vs. non-recursive storyboard variant) or the cinematic language injection (e.g., vs. generic shot prompts). Without these controls, the central attribution of improvements to the new agent's specific mechanisms remains untested and load-bearing for the outperformance claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. These observations highlight opportunities to improve clarity and empirical rigor. We address each major comment point by point below, indicating the revisions we will make to the next version of the paper.

read point-by-point responses

Referee: Abstract: The claim that the approach 'consistently outperforms existing baselines' in narrative consistency, dynamic expressiveness, and perceived film quality is not accompanied by any specific metrics, baseline names, dataset details, or ablation results. This makes it impossible to verify whether gains are attributable to the Cinematography Shot Agent's recursive storyboard generation and cinematic language injection rather than confounding factors such as base model choice or overall prompt complexity.

Authors: We agree that the abstract, as currently written, is too high-level to allow immediate verification of the claims. The Experiments section of the full manuscript already reports the specific metrics (narrative consistency, dynamic expressiveness, and film quality scores), baseline names, dataset details, and controls that use identical base video-generation models across comparisons. To address the referee's concern directly, we will revise the abstract to concisely summarize these key quantitative results and explicitly note that the same underlying models were used to isolate the contributions of the Cinematography Shot Agent. revision: yes
Referee: Experiments section (inferred from abstract's reference to quantitative/qualitative results): No ablation studies are described that isolate the recursive storyboard generation (e.g., full model vs. non-recursive storyboard variant) or the cinematic language injection (e.g., vs. generic shot prompts). Without these controls, the central attribution of improvements to the new agent's specific mechanisms remains untested and load-bearing for the outperformance claim.

Authors: We acknowledge that the current manuscript does not present dedicated ablation studies that isolate the recursive storyboard generation and the cinematic language injection components. While the existing baseline comparisons and qualitative analyses provide supporting evidence, explicit ablations would strengthen the causal attribution. In the revised manuscript we will add two targeted ablation experiments: (1) a non-recursive storyboard variant and (2) a generic-prompt variant without cinematic language injection. Both will be evaluated under the same metrics, datasets, and base models as the full model to directly test the contribution of each mechanism. revision: yes

Circularity Check

0 steps flagged

No circularity: new agent components and evaluation results are independent of inputs

full rationale

The paper describes an extension of prior multi-agent video generation pipelines by adding a Cinematography Shot Agent that performs recursive storyboard generation and cinematic language injection. These are presented as novel mechanisms whose value is measured via separate quantitative and qualitative experiments on narrative consistency, dynamic expressiveness, and film quality. No equations, fitted parameters, or self-referential definitions appear; the outperformance claim is not reduced to a renaming or re-use of the input baselines by construction. The derivation chain therefore remains self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that multi-agent systems can be extended with cinematography-specific mechanisms to improve filmic qualities, plus the new agent component whose effectiveness is asserted via experiments.

axioms (1)

domain assumption Established agentic pipelines provide a sufficient base that can be improved by adding domain-specific agents for cinematography and narrative continuity
The paper states it builds upon established agentic pipelines without independent validation of the base assumption in the abstract.

invented entities (1)

Cinematography Shot Agent no independent evidence
purpose: To perform recursive storyboard generation and inject cinematic language for improved shot design and continuity
Newly introduced component that is central to the proposed framework and not present in the baselines.

pith-pipeline@v0.9.0 · 5422 in / 1276 out tokens · 58744 ms · 2026-05-10T16:47:33.366861+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

21 extracted references · 10 canonical work pages · 4 internal anchors

[1]

Wan: Open and Advanced Large-Scale Video Generative Models

Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, et al., “Wan: Open and advanced large-scale video generative models,”arXiv preprint arXiv:2503.20314, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[2]

ModelScope Text-to-Video Technical Report

Jiuniu Wang, Hangjie Yuan, Dayou Chen, Yingya Zhang, Xiang Wang, and Shiwei Zhang, “Modelscope text-to-video technical report,”arXiv preprint arXiv:2308.06571, 2023

work page internal anchor Pith review arXiv 2023
[3]

HunyuanVideo: A Systematic Framework For Large Video Generative Models

Weijie Kong, Qi Tian, Zijian Zhang, Rox Min, Zuozhuo Dai, Jin Zhou, Jiangfeng Xiong, Xin Li, Bo Wu, Jianwei Zhang, et al., “Hunyuanvideo: A systematic framework for large video generative models,”arXiv preprint arXiv:2412.03603, 2024

work page internal anchor Pith review Pith/arXiv arXiv 2024
[4]

Starvid: Enhancing semantic alignment in video diffusion models via spatial and syntactic guided attention refocusing,

Yuanhang Li, Qi Mao, Lan Chen, Zhen Fang, Lei Tian, Xinyan Xiao, Libiao Jin, and Hua Wu, “Starvid: Enhancing semantic alignment in video diffusion models via spatial and syntactic guided attention refocusing,”arXiv preprint arXiv:2409.15259, 2024

work page arXiv 2024
[5]

Captain cinema: Towards short movie generation,

Junfei Xiao, Ceyuan Yang, Lvmin Zhang, Shengqu Cai, Yang Zhao, Yuwei Guo, Gordon Wetzstein, Maneesh Agrawala, Alan Yuille, and Lu Jiang, “Captain cinema: Towards short movie generation,”arXiv preprint arXiv:2507.18634, 2025

work page arXiv 2025
[6]

Multi-agent systems: A survey,

Ali Dorri, Salil S Kanhere, and Raja Jurdak, “Multi-agent systems: A survey,”IEEE Access, 2018

2018
[7]

A survey on evaluation of large language models,

Yupeng Chang, Xu Wang, Jindong Wang, Yuan Wu, Linyi Yang, Kaijie Zhu, Hao Chen, Xiaoyuan Yi, Cunxiang Wang, Yidong Wang, et al., “A survey on evaluation of large language models,”TIST, 2024

2024
[8]

Anim-director: A large multimodal model powered agent for controllable animation video generation,

Yunxin Li, Haoyuan Shi, Baotian Hu, Longyue Wang, Jiashun Zhu, Jinyi Xu, Zhen Zhao, and Min Zhang, “Anim-director: A large multimodal model powered agent for controllable animation video generation,” in SIGGRAPH Asia, 2024

2024
[9]

Dreamstory: Open-domain story visualization by llm-guided multi- subject consistent diffusion,

Huiguo He, Huan Yang, Zixi Tuo, Yuan Zhou, Qiuyue Wang, Yuhang Zhang, Zeyu Liu, Wenhao Huang, Hongyang Chao, and Jian Yin, “Dreamstory: Open-domain story visualization by llm-guided multi- subject consistent diffusion,”PAMI, 2025

2025
[10]

Videogen-of-thought: Step-by-step generating multi-shot video with minimal manual intervention.arXiv preprint arXiv:2412.02259, 2024

Mingzhe Zheng, Yongqi Xu, Haojian Huang, Xuran Ma, Yexin Liu, Wenjie Shu, Yatian Pang, Feilong Tang, et al., “Videogen-of-thought: Step-by-step generating multi-shot video with minimal manual interven- tion,”arXiv preprint arXiv:2412.02259, 2024

work page arXiv 2024
[11]

arXiv preprint arXiv:2503.07314 (2025)

Weijia Wu, Zeyu Zhu, and Mike Zheng Shou, “Automated movie gen- eration via multi-agent cot planning,”arXiv preprint arXiv:2503.07314, 2025

work page arXiv 2025
[12]

Chain-of-thought prompting elicits reasoning in large language models,

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al., “Chain-of-thought prompting elicits reasoning in large language models,” inNeurIPS, 2022

2022
[13]

Lora: Low-rank adaptation of large language models.,

Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al., “Lora: Low-rank adaptation of large language models.,” inICLR, 2022

2022
[14]

“Gpt-4o,” Accessed May 13, 2024 [Online] https://openai.com/index/ hello-gpt-4o/

2024
[15]

Qwen3 Technical Report

An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al., “Qwen3 technical report,”arXiv preprint arXiv:2505.09388, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[16]

Shotbench: Expert-level cinematic understanding in vision-language models,

Hongbo Liu, Jingwen He, Yi Jin, Dian Zheng, Yuhao Dong, Fan Zhang, Ziqi Huang, Yinan He, Yangguang Li, Weichao Chen, et al., “Shotbench: Expert-level cinematic understanding in vision-language models,”arXiv preprint arXiv:2506.21356, 2025

work page arXiv 2025
[17]

Magref: Masked guidance for any-reference video generation,

Yufan Deng, Xun Guo, Yuanyang Yin, Jacob Zhiyuan Fang, Yiding Yang, Yizhi Wang, Shenghai Yuan, Angtian Wang, Bo Liu, Haibin Huang, et al., “Magref: Masked guidance for any-reference video generation,”arXiv preprint arXiv:2505.23742, 2025

work page arXiv 2025
[18]

Flux. 1 kontext: Flow matching for in-context image generation and editing in latent space,

Stephen Batifol, Andreas Blattmann, Frederic Boesel, Saksham Consul, Cyril Diagne, Tim Dockhorn, Jack English, Zion English, Patrick Esser, et al., “Flux. 1 kontext: Flow matching for in-context image generation and editing in latent space,”arXiv e-prints, pp. arXiv–2506, 2025

2025
[19]

VBench: Compre- hensive benchmark suite for video generative models,

Ziqi Huang, Yinan He, Jiashuo Yu, Fan Zhang, Chenyang Si, Yuming Jiang, Yuanhan Zhang, Tianxing Wu, Qingyang Jin, Nattapol Chanpaisit, Yaohui Wang, Xinyuan Chen, Limin Wang, et al., “VBench: Compre- hensive benchmark suite for video generative models,” inCVPR, 2024

2024
[20]

Learning transferable visual models from natural language supervision,

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al., “Learning transferable visual models from natural language supervision,” inICML, 2021, pp. 8748–8763

2021
[21]

Gemini: a family of highly capable multimodal models,

Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M Dai, Anja Hauth, Katie Millican, et al., “Gemini: a family of highly capable multimodal models,” Tech. Rep., 2023. Camera Artist: A Multi-Agent Framework for Cinematic Language Storytelling Video Generation Supplementary Material APPENDIX...

2023