Pith. sign in

REVIEW 4 major objections 8 minor 82 references

MapStory: Prototyping Editable Map Animations with LLM Agents

T0 review · 4 major / 8 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A dual-agent LLM system turns written scripts into editable map animations whose zooms, highlights, and routes are grounded in real geospatial data.

desk verdict Solid UIST-style systems paper with a genuinely new dual-agent architecture and a useful taxonomy; the factual-grounding claim is load-bearing and under-evidenced, but the editability value holds up. read the letter →

arxiv 2505.21966 v2 pith:RQ4OUWP4 submitted 2025-05-28 cs.HC cs.AIcs.CLcs.MM

Aditya Gunturu , Ben Pearman , Keiichi Ihara , Morteza Faraji , Bryan Wang , Rubaiat Habib Kazi , Ryo Suzuki This is my paper
classification cs.HCcs.AIcs.CLcs.MM
keywords text-to-animationmap-basedstorytellingLLMagentsgeospatialdatagroundingeditableanimationtimelinehuman-AIcollaborationcreativesupporttoolsGeoJSON
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

MapStory is a system that lets a user write a short script in plain English and receive an editable map animation: a sequence of camera moves, highlights, and animated elements arranged on a timeline. The paper argues that a two-agent LLM architecture can perform both halves of what previously took professional animators days: breaking a story into animation beats, and researching the real geographic coordinates, boundaries, and routes those beats need. If the claim holds, map-centric storytelling stops being a specialist skill and becomes a prototyping task available to teachers, journalists, travelers, and casual creators. The paper supports the claim with a formative taxonomy of 200 videos, a usability study with twelve novices, interviews with five professional animators, and a technical check on twenty prompts.

What carries the argument

The load-bearing object is the two-agent LLM architecture, combined with a fixed vocabulary of primitive map-animation blocks. A scene breakdown agent converts the script into a JSON sequence of blocks with names, start and end times, and initial parameters, following planning and chain-of-thought reasoning. A researcher agent with web search then takes each block, validates its parameters, and calls a single function-calling API to fetch or construct GeoJSON via a geocoding service, using one of five actions: query, combination, modification, reduction, or generation. The timeline editor renders those blocks as editable modules whose parameters can be changed without disturbing other blocks. The taxonomy of primitives comes from an analysis of 200 map animation videos, and the agent prompts encode heuristics such as always preceding a highlight with a camera move and scheduling modules in non-overlapping intervals.

What would settle it

Run the system on a fixed set of twenty prompts with known ground-truth geography, including historical queries like the Roman Empire and ambiguous names like Avon River, and compare every returned polygon, route, and point against authoritative sources; the central claim weakens if the share of modules accurate within one kilometer falls well below the paper's reported rates, or if users must correct most blocks before the animation is usable.

Watch

Extended reading notes

Core claim

The central claim is that natural-language scripts can be converted into editable, geospatially grounded map animations through a dual-agent LLM pipeline, and that this conversion is fast, accurate enough for prototyping, and controllable at every stage. The scene breakdown agent parses the script into modular primitive blocks: area, line, and point highlights; zoom, translate, and orbit camera moves; and animated routes, spatial transitions, and auxiliary motion. The researcher agent then grounds each block in real geospatial data by querying a geocoding service with web search, using actions that query, combine, modify, reduce, or generate GeoJSON. Users can rewrite the script, reorder blocks, edit parameters and timing in a timeline editor, or chat with the researcher agent to refine data. The paper's evaluations report that novices rated 79.82% of generated modules as accurate, with the guided task reaching 89.47% before corrections, and that experts saw the tool saving roughly two days of a five-day production schedule.

Load-bearing premise

The system's promise of factual grounding rests on the researcher agent returning accurate, current boundaries, paths, and coordinates from web search and geocoding; if geocoding is materially wrong, the animation is easy to make but not trustworthy.

Editorial extensions

If this is right

  • Novices can prototype map animations in minutes rather than days, since the script-to-scene breakdown removes the need to hand-place keyframes.
  • The researcher agent collapses the separate research phase into the animation tool, so last-minute factual changes update the scene instead of forcing manual rework.
  • Because each animation block is an editable JSON module with independent parameters, users can reorder, delete, or tweak one beat without disturbing the rest of the timeline.
  • Professional animators gain a pre-production layer that generates draft animations quickly enough to explore multiple storylines in parallel before committing to full-resolution rendering.
  • The underlying LLM's capability bounds the whole system: the paper's bottleneck analysis says a single prompt can produce at most about 54 animation blocks under current token limits, so richer stories require iterative prompting.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The scene breakdown could function as a portable edit-decision list: if the block format is standardized, the same text-driven timeline might be imported into other animation tools, extending the paper's expert suggestion that the breakdown be saved and reused.
  • The researcher agent's five GeoJSON actions suggest a general pattern for grounding LLM output in structured data; the same query-combine-modify-reduce-generate toolbox could apply to other domains where language must be tied to verified records, not just maps.
  • The paper's reported accuracy (about 80% subjective on open-ended tasks, with geocoding errors typically within 1 km) implies that the tool is best suited for concept and pitch work, where slight boundary imprecision is tolerable, rather than publication-grade cartography.
  • A testable extension would feed the same prompts to a verification agent that cross-checks each GeoJSON against a second geocoder, measuring how much accuracy improves; the paper's future-work section already floats a multi-round verification architecture.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 8 minor

Summary. MapStory is an LLM-powered authoring tool that converts natural-language scripts into editable map animations. The system uses a scene breakdown agent to decompose a script into modular primitives (camera movements, highlights, animated elements) and a researcher agent to fetch geospatial data through web search and the Nominatim API. The design is grounded in a formative analysis of 200 map animation videos and interviews with professional animators. The evaluation comprises a technical evaluation of 20 prompts across four LLMs, a usability study with 12 participants (SUS 83.7), and expert interviews with 5 professionals. The paper claims that MapStory facilitates faster iteration, creative exploration, and lower barriers to map animation creation.

Significance. The system is a plausible and well-motivated contribution to accessible map animation authoring. The primary strengths are the modular primitive design grounded in a 200-video corpus, the explicit human-in-the-loop architecture, and the candid documentation of failure cases such as historical boundary queries and hallucination. The central limitation is the evaluation of factual grounding: the accuracy claims rely on author grading of 20 prompts and on novice participants' self-reports, without objective ground truth or a baseline. If the accuracy evidence is strengthened or the claims modestly scaled back, the contribution stands as a valuable prototyping tool rather than as a verified factual-grounding system.

major comments (4)
  1. [Section 7 and Abstract] The claim that the researcher agent 'accurately queries geospatial information' rests on the authors' own grading of 20 GPT-4.5-generated prompts with no objective ground truth, no baseline, and no inter-rater reliability (Section 7.1). The grading criteria are permissive: animated routes are considered accurate if within 1 km of the real route and 'physically plausible.' The paper itself concedes (Section 7.2 and Section 10) that historical queries such as the Roman Empire return only present-day fragments and that geocoding errors within about 1 km occur (Section 8.2). Because factual grounding is an explicit feature of the design (abstract and Section 6.3), this evidence is insufficient to support the claim as written; the authors should either temper the claim to 'supports geospatial lookups with human verification' or add an objective evaluation against authoritative geospatial data with independent annotation.
  2. [Section 8.2] The 'subjective accuracy rate' of 79.82% is computed from participants' self-reports of which modules they believed matched their prompts (Section 8.1), not from verification of the geospatial content. Novice participants are unlikely to detect incorrect boundaries, mislocated points, or historically anachronistic regions, so this metric provides little support for the correctness component of the paper's claims. The statement that Task A accuracy reached 100% 'after the user changes' further indicates that the metric reflects user satisfaction with their own edits rather than objective accuracy.
  3. [Section 7.1] The decision to omit a baseline is justified by an undocumented test in which Sora 'failed completely' on five prompts. No description of the prompts, outputs, or failure criteria is given, so the reader cannot assess the comparison. The technical evaluation should include a clearly specified baseline (or state that no comparison is being made) and should report the agreement between the authors' grading and an independent annotation, given that all accuracy judgments are subjective.
  4. [Section 6.3] The researcher agent's tool architecture is underspecified. The system is said to use Perplexity's sonar-pro 'with web-search and reasoning capability' and also that 'each research agent can only access one tool: a function call adhering to the OpenAI Function Calling Protocol.' It is unclear whether the web search is an internal capability of sonar-pro or a separate tool call, and how Nominatim lookups are combined with web-sourced GeoJSON for historical or non-queryable regions. This ambiguity impedes reproduction and should be clarified with a concrete example from the walkthrough.
minor comments (8)
  1. [Section 1] In the enumerated limitations near the end of the Introduction, the phrase 'challenges with AI hallucination affecting factual accuracy the need to verify produced visualizations' is missing a conjunction; please rephrase.
  2. [Section 6.3] The sentence beginning 'Given a user𝑠𝑐𝑟𝑖𝑝𝑡' contains a Unicode control character; replace it with 'Given a user script'.
  3. [Section 6.3.1] The heading 'GeoJSON Retrival' should be 'GeoJSON Retrieval'.
  4. [Section 7.1] The sentence 'We did not include a baseline because we simply tested our system with text-to-video generation with Sora but with a simple test of 5 prompts, we immediately realized that this approach failed completely' is grammatically tangled and should be rewritten for clarity.
  5. [Section 7.2] The statement that 'GPT-4.5 took less time to produce an output for the scene breakdown' appears to contradict the preceding sentence that GPT-4.5 took the longest overall; please clarify what comparison is being made.
  6. [Section 8.2] The phrase 'After the user changes, the Task A success rate was 100%' is ambiguous; specify what the user changed and how the recalculated success rate was derived.
  7. [References] Reference [67] is incomplete: the title ends mid-phrase ('Prompting LLMs to Solve Complex Tasks: A.') and the publication venue is missing; please complete the citation.
  8. [Section 10] The statement 'Our agent design is also limited by a design features two agents' contains a grammatical error; it should be 'limited by a design that features two agents.'

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; MapStory's claims are supported by user studies and external geospatial checks, not by self-referential derivation.

full rationale

MapStory is a systems/HCI paper: its central claims concern usability, iteration speed, creative exploration, and prototyping support, and these are assessed through a usability study (N=12), expert interviews (N=5), and a technical evaluation of 20 prompts. There is no derived equation, fitted parameter, or formal model whose output is definitionally equal to its input. The technical evaluation in Section 7.1 grades accuracy by author judgment with no objective ground truth ('We evaluated accuracy based on our judgment as no standardized or objective ground truth was available for these types of outputs'), and Section 8.2 reports a 79.82% 'subjective accuracy' from participants' self-reports. These are evidence-quality limitations, not circularity, because the system's outputs are still being compared against external geographic reality and user experience rather than being constructed to match the evaluation criterion. The paper explicitly concedes failure cases, including historical queries such as the Roman Empire and geocoding errors of about 1 km, which would be surprising if the claims were definitionally forced. Self-citations in the related work (e.g., Augmented Physics, CrossTalk, LAVE) are contextual and not load-bearing for any of the paper's contributions. The formative taxonomy from 200 videos and the expert-informed design goals are empirical inputs, not outputs of the system. Overall, no load-bearing circular step reduces MapStory's contributions to its own premises.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No numeric free parameters or new physical entities are introduced. The central design rests on assumptions about LLM planning, geodata accuracy, taxonomy representativeness, and the transfer of expert workflows to novices, each of which is plausible but not independently established.

assumptions (4)
  • domain assumption The two-agent LLM architecture can decompose scripts into correct primitive animation modules and geocode references accurately enough for usable output.
    Assumed throughout Sections 5 and 6; tested only in Section 7 with 20 author-graded prompts and no objective ground truth.
  • domain assumption The taxonomy of map animation primitives derived from 200 YouTube videos is representative of creator needs.
    Section 4.1 describes thematic coding by two authors and group consensus, but no inter-rater reliability or independent validation is reported.
  • domain assumption OpenStreetMap/Nominatim and LLM web search provide sufficiently accurate and current geodata for factual animation.
    Invoked in Section 6.3; the paper itself acknowledges limitations for historical boundaries and occasional 1 km geocoding errors.
  • domain assumption Expert-derived design guidance transfers to novice users (cognitive apprenticeship).
    Stated in Sections 3 and 8.1 based on prior work such as K-Sketch and Draco; not directly tested in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MapStory: Prototyping Editable Map Animations with LLM Agents." pith.science (2026). https://pith.science/paper/RQ4OUWP4

@misc{pith2026250521966,
  author       = {Pith},
  title        = {Pith review of: MapStory: Prototyping Editable Map Animations with LLM Agents},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RQ4OUWP4}},
  note         = {Machine review of arXiv:2505.21966}
}
read the original abstract

We introduce MapStory, an LLM-powered animation prototyping tool that generates editable map animation sequences directly from natural language text by leveraging a dual-agent LLM architecture. Given a user written script, MapStory automatically produces a scene breakdown, which decomposes the text into key map animation primitives such as camera movements, visual highlights, and animated elements. Our system includes a researcher agent that accurately queries geospatial information by leveraging an LLM with web search, enabling automatic extraction of relevant regions, paths, and coordinates while allowing users to edit and query for changes or additional information to refine the results. Additionally, users can fine-tune parameters of these primitive blocks through an interactive timeline editor. We detail the system's design and architecture, informed by formative interviews with professional animators and by an analysis of 200 existing map animation videos. Our evaluation, which includes expert interviews (N=5) and a usability study (N=12), demonstrates that MapStory enables users to create map animations with ease, facilitates faster iteration, encourages creative exploration, and lowers barriers to creating map-centric stories.

Figures

Figures reproduced from arXiv: 2505.21966 by the authors.

Figure 1
Figure 1. MapStory is an LLM-powered tool for prototyping map-centric animations from natural language scripts. Users write [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Popular online animated map videos explain knowledge to a general audience in a highly engaging manner, influencing [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Workflow of a map animator. The animator first [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Common map augmentation and animation techniques based on content analysis of 200 map-centric videos (See list [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: MapStory System Walkthrough. Highlights. Highlights are a primary method of emphasizing spe￾cific geographic areas of interest on a map. They can be imple￾mented as areas, lines, or points: • Area highlights are used to visually emphasize regions or zones by shading, c…
Figure 6
Figure 6. Figure 6: Map Stylization 5https://nominatim.openstreetmap.org/ui/search.html 6https://docs.mapbox.com/api/maps/styles/ 8 [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: Camera controls: 1) zoom, 2) orbit, and 3) transla [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Animated elements: 1) routes, 2) auxiliary motion, [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: MapStory LLM Architecture issues or clarity [67]. We chose the two main tasks of creating map animations as found in our formative explorations: an animation planning stage and a research stage. Our architecture features two primary LLM agents: a scene breakdown agent…
Figure 11
Figure 11. Figure 11: Results of accuracy across four LLM models. [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 12
Figure 12. Figure 12: Results of processing time (in seconds) across four [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 14
Figure 14. Figure 14: Results of the perceived usefulness of each work [PITH_FULL_IMAGE:figures/full_fig_p013_14.png]
Figure 13
Figure 13. Figure 13: Results of the questionnaire for each system fea [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

82 extracted references · 61 canonical work pages

  1. [1]

    TravelAnimator: Accessible Map Animation Tool

    [n.d.]. TravelAnimator: Accessible Map Animation Tool. https://travelanimator. com/. Accessed: 2025-07-15

    2025

  2. [2]

    Catalan Atlas

    1375. Catalan Atlas. https://en.wikipedia.org/wiki/Catalan_Atlas

  3. [3]

    Carte figurative des pertes successives en hommes de l’Armée Française dans la campagne de Russie 1812–1813

    1869. Carte figurative des pertes successives en hommes de l’Armée Française dans la campagne de Russie 1812–1813. https://en.wikipedia.org/wiki/Charles_ Joseph_Minard. 15 UIST ’25, September 28-October 1, 2025, Busan, Republic of Korea Gunturu, et al

    work page 2025

  4. [4]

    2024. Veo 2. https://deepmind.google/models/veo/

    work page 2024

  5. [5]

    AEScripts. 2025. GeoLayers: The Ultimate Map and Data Visualization Plugin for After Effects. https://aescripts.com/geolayers/ Accessed: 2025-03-12

    work page 2025

  6. [6]

    Tyler Angert, Miroslav Suzara, Jenny Han, Christopher Pondoc, and Hariharan Subramonyam. 2023. Spellburst: A node-based interface for exploratory creative coding with natural language prompts. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology . 1–22

    2023

  7. [7]

    Ian Arawjo, Priyan Vaithilingam, Martin Wattenberg, and Elena Glassman. 2023. ChainForge: An open-source visual programming environment for prompt en- gineering. In Adjunct proceedings of the 36th annual ACM symposium on user interface software and technology . 1–3

    work page 2023

  8. [8]

    Jan Henry Belz, Lina Madlin Weilke, Anton Winter, Philipp Hallgarten, Enrico Rukzio, and Tobias Grosse-Puppendahl. 2024. Story-Driven: Exploring the Im- pact of Providing Real-time Context Information on Automated Storytelling. In Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology. 1–15

    work page 2024

Show all 82 references
  1. [9]

    Stephen Brade, Bryan Wang, Mauricio Sousa, Sageev Oore, and Tovi Gross- man. 2023. Promptify: Text-to-image generation through interactive prompt exploration with large language models. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology . 1–14

    2023

  2. [10]

    John Brooke et al. 1996. SUS-A quick and dirty usability scale.Usability evaluation in industry 189, 194 (1996), 4–7

    1996

  3. [11]

    Tim Brooks, Bill Peebles, Connor Holmes, Will DePue, Yufei Guo, Li Jing, David Schnurr, Joe Taylor, Troy Luhman, Eric Luhman, et al. 2024. Video generation models as world simulators. OpenAI Blog 1 (2024), 8

    2024

  4. [12]

    Yining Cao, Jane L E, Zhutian Chen, and Haijun Xia. 2023. DataParticles: Block- based and language-oriented authoring of animated unit visualizations. In Pro- ceedings of the 2023 CHI Conference on Human Factors in Computing Systems . 1–15

    2023

  5. [13]

    Mengyu Chen, Marko Peljhan, and Misha Sra. 2021. Entanglevr: A visual pro- gramming interface for virtual reality interactive scene generation. InProceedings of the 27th ACM symposium on virtual reality software and technology . 1–6

    2021

  6. [14]

    Erin Cherry and Celine Latulipe. 2014. Quantifying the creativity support of digital tools through the creativity support index.ACM Transactions on Computer- Human Interaction (TOCHI) 21, 4 (2014), 1–25

    2014

  7. [15]

    John Joon Young Chung, Wooseok Kim, Kang Min Yoo, Hwaran Lee, Eytan Adar, and Minsuk Chang. 2022. TaleBrush: Sketching stories with generative pretrained language models. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems. 1–19

    2022

  8. [16]

    Richard C Davis, Brien Colwell, and James A Landay. 2008. K-sketch: a’kinetic’sketch pad for novice animators. In Proceedings of the SIGCHI Con- ference on Human Factors in Computing Systems . 413–422

    2008

  9. [17]

    Samantha Dawson. 2024. OpenAI launches experimental GPT-4o Long Output model with 16 × token capacity. VentureBeat (15 Aug. 2024). https://venturebeat.com/ai/openai-launches-experimental-gpt-4o-long- output-model-with-16x-token-capacity/

    2024

  10. [18]

    Vanessa P Dennen and Kerry J Burner. 2008. The cognitive apprenticeship model in educational practice. In Handbook of research on educational communications and technology. Routledge, 425–439

    2008

  11. [19]

    Tong Gao, Jessica R Hullman, Eytan Adar, Brent Hecht, and Nicholas Diakopoulos

  12. [20]

    Google. 2025. Veo 2 - Google DeepMind. https://deepmind.google/technologies/ veo/veo-2/ [Online; accessed 2025-03-31]

    2025

  13. [21]

    Aditya Gunturu, Yi Wen, Nandi Zhang, Jarin Thundathil, Rubaiat Habib Kazi, and Ryo Suzuki. 2024. Augmented Physics: Creating Interactive and Embedded Physics Simulations from Static Textbook Diagrams. In Proceedings of the 37th Annual ACM Symposium on User Interface Software a...

    2024

  14. [22]

    Anton Gustafsson, John Bichard, Liselott Brunnberg, Oskar Juhlin, and Marco Combetto. 2006. Believable environments: generating interactive storytelling in vast location-based pervasive games. In Proceedings of the 2006 ACM SIGCHI international conference on Advances in comput...

    2006

  15. [23]

    Geo History. 2021. History of the Major Trade Routes - Summary on a Map. https://www.youtube.com/watch?v=Q7xp1-VvtZ0 Accessed: 2025-01-06

    2021

  16. [24]

    Bernd Huber, Hijung Valentina Shin, Bryan Russell, Oliver Wang, and Gautham J Mysore. 2019. B-script: Transcript-based b-roll video editing with recommenda- tions. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. 1–11

    2019

  17. [25]

    Amir Jahanlou and Parmit K Chilana. 2022. Katika: An end-to-end system for authoring amateur explainer motion graphics videos. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems . 1–14

    2022

  18. [26]

    Peiling Jiang, Jude Rayan, Steven P Dow, and Haijun Xia. 2023. Graphologue: Exploring large language model responses with interactive diagrams. In Proceed- ings of the 36th annual ACM symposium on user interface software and technology . 1–20

    2023

  19. [27]

    Rubaiat Habib Kazi, Fanny Chevalier, Tovi Grossman, and George Fitzmaurice

  20. [28]

    Rubaiat Habib Kazi, Fanny Chevalier, Tovi Grossman, Shengdong Zhao, and George Fitzmaurice. 2014. Draco: Bringing life to illustrations. InCHI’14 Extended Abstracts on Human Factors in Computing Systems . 579–582

    2014

  21. [29]

    In Proceedings of the 27th annual ACM symposium on User interface software and technology

    Kitty: sketching dynamic and interactive illustrations. In Proceedings of the 27th annual ACM symposium on User interface software and technology . 395–405

  22. [30]

    Brian Felipe Keith Norambuena and Tanushree Mitra. 2021. Narrative maps: An algorithmic approach to represent and extract information narratives.Proceedings of the ACM on Human-Computer Interaction 4, CSCW3 (2021), 1–33

    2021

  23. [31]

    Rubaiat Habib Kazi, Tovi Grossman, Nobuyuki Umetani, and George Fitzmaurice

  24. [32]

    Mackenzie Leake, Abe Davis, Anh Truong, and Maneesh Agrawala. 2017. Com- putational video editing for dialogue-driven scenes. ACM Trans. Graph. 36, 4 (2017), 130–1

    2017

  25. [33]

    Mackenzie Leake and Wilmot Li. 2024. ChunkyEdit: Text-first video interview editing via chunking. In Proceedings of the CHI Conference on Human Factors in Computing Systems. 1–16

    2024

  26. [34]

    Knowledgia. 2022. How did The Silk Road Actually Work? https://www.youtube. com/watch?v=J-pfeFbssMw Accessed: 2025-01-06

    2022

  27. [35]

    Wanwan Li, Changyang Li, Minyoung Kim, Haikun Huang, and Lap-Fai Yu. 2023. Location-aware adaptation of augmented reality narratives. In Proceedings of the 2023 CHI conference on human factors in computing systems . 1–15

    2023

  28. [36]

    Wenchao Li, Zhan Wang, Yun Wang, Di Weng, Liwenhan Xie, Siming Chen, Haidong Zhang, and Huamin Qu. 2023. GeoCamera: Telling stories in geographic visualizations with camera movements. In Proceedings of the 2023 CHI conference on human factors in computing systems . 1–15

    2023

  29. [37]

    Fan Lei, Yuxin Ma, A Stewart Fotheringham, Elizabeth A Mack, Ziqi Li, Mehak Sachdeva, Sarah Bardin, and Ross Maciejewski. 2023. GeoExplainer: A visual analytics framework for spatial modeling contextualization and report generation. IEEE Transactions on Visualization and Compu...

    2023

  30. [38]

    Real life Core. 2019. Why 50% of Canadians Live South of This Line. https: //www.youtube.com/watch?v=DFJAgb7dn78 Accessed: 2025-01-06

    2019

  31. [39]

    Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang

    Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2023. Lost in the Middle: How Language Models Use Long Contexts. arXiv:2307.03172 [cs.CL] https://arxiv.org/abs/2307.03172

    2023 arXiv

  32. [40]

    Jian Liao, Adnan Karim, Shivesh Singh Jadon, Rubaiat Habib Kazi, and Ryo Suzuki

  33. [41]

    Patrik Lundblad and Mikael Jern. 2013. Geovisual analytics and storytelling using HTML5. In 2013 17th International Conference on Information Visualisation . IEEE, 263–271

    2013

  34. [42]

    Jiaju Ma, Li-Yi Wei, and Rubaiat Habib Kazi. 2022. A layered authoring tool for stylized 3d animations. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems . 1–14

    2022

  35. [43]

    Rohin Manvi, Samar Khanna, Gengchen Mai, Marshall Burke, David Lobell, and Stefano Ermon. 2023. Geollm: Extracting geospatial knowledge from large language models. arXiv preprint arXiv:2310.06213 (2023)

    2023 arXiv

  36. [44]

    Vivian Liu, Rubaiat Habib Kazi, Li-Yi Wei, Matthew Fisher, Timothy Langlois, Seth Walker, and Lydia Chilton. 2024. LogoMotion: Visually Grounded Code Generation for Content-Aware Animation.arXiv preprint arXiv:2405.07065 (2024)

    2024 arXiv

  37. [45]

    Tomer Moscovich and John F Hughes. 2001. Animation sketching: An approach to accessible animation. Unpublished Master’s Thesis, CS Department, Brown University 5 (2001)

    2001

  38. [46]

    Life of a Scrambler. 2025. How I Make My Maps. https://www.youtube.com/ watch?v=GsojLuJpe_0

    2025

  39. [47]

    OpenAI. 2023. Function Calling and Other API Updates. https://openai.com/ blog/functioncalling-and-other-api-updates. Accessed: 2023-XX-XX

    2023

  40. [48]

    Damien Masson, Sylvain Malacria, Géry Casiez, and Daniel Vogel. 2024. Direct- gpt: A direct manipulation interface to interact with large language models. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems . 1–16

    2024

  41. [49]

    Pika. 2025. Pika. https://pika.art/ [Online; accessed 2025-03-31]

    2025

  42. [50]

    Jonathan Roberts, Timo Lüddecke, Sowmen Das, Kai Han, and Samuel Albanie

  43. [51]

    Karl Toby Rosenberg, Rubaiat Habib Kazi, Li-Yi Wei, Haijun Xia, and Ken Perlin

  44. [52]

    OpenAI. 2025. Sora. https://openai.com/sora/. [Online; accessed 28-March-2025]

    2025

  45. [53]

    Kadek Ananta Satriadi, Barrett Ens, Tobias Czauderna, Maxime Cordeil, and Bern- hard Jenny. 2021. Quantitative data visualisation on virtual globes. InProceedings of the 2021 CHI Conference on Human Factors in Computing Systems . 1–14

    2021

  46. [54]

    Kadek Ananta Satriadi, Jim Smiley, Barrett Ens, Maxime Cordeil, Tobias Czaud- erna, Benjamin Lee, Ying Yang, Tim Dwyer, and Bernhard Jenny. 2022. Tangible 16 MapStory: Prototyping Editable Map Animations with LLM Agents UIST ’25, September 28-October 1, 2025, Busan, Republic o...

    2022

  47. [55]

    Vidya Setlur, Sarah E Battersby, Melanie Tory, Rich Gossweiler, and Angel X Chang. 2016. Eviza: A natural language interface for visual analysis. InProceedings of the 29th annual symposium on user interface software and technology . 365–377

    2016

  48. [56]

    Jae-Eun Shin and Woontack Woo. 2023. How space is told: linking trajectory, narrative, and intent in augmented reality storytelling for cultural heritage sites. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems . 1–14

    2023

  49. [57]

    Paul Sonnentag, Alexander Obenauer, and Geoffrey Litt. 2023. Embark: Dynamic Documents as Personal Software. In LIVE 2023: The Ninth Workshop on Live Programming. Cascais, Portugal. https://www.inkandswitch.com/embark/

    2023

  50. [58]

    Runway AI

    Inc. Runway AI. 2025. Runway | Tools for human imagination. https://runwayml. com/ [Online; accessed 2025-03-31]

    2025

  51. [59]

    Bekzat Tilekbay, Saelyne Yang, Michal Adam Lewkowicz, Alex Suryapranata, and Juho Kim. 2024. ExpressEdit: Video Editing with Natural Language and Sketching. In Proceedings of the 29th International Conference on Intelligent User Interfaces. 515–536

    2024

  52. [60]

    Anh Truong, Floraine Berthouzoz, Wilmot Li, and Maneesh Agrawala. 2016. Quickcut: An interactive tool for editing narrated video. In Proceedings of the 29th Annual Symposium on User Interface Software and Technology . 497–507

    2016

  53. [61]

    Tiffany Tseng, Ruijia Cheng, and Jeffrey Nichols. 2024. Keyframer: Empowering animation design using large language models. arXiv preprint arXiv:2402.06071 (2024)

    2024 arXiv

  54. [62]

    Edward R Tufte. 1991. Envisioning information. Optometry and Vision Science 68, 4 (1991), 322–324

    1991

  55. [63]

    1983.The visual display of quantitative information

    Edward R Tufte and Peter R Graves-Morris. 1983.The visual display of quantitative information. Vol. 2. Graphics press Cheshire, CT

    1983

  56. [64]

    Hariharan Subramonyam, Wilmot Li, Eytan Adar, and Mira Dontcheva. 2018. Taketoons: Script-driven performance animation. InProceedings of the 31st Annual ACM Symposium on User Interface Software and Technology . 663–674

    2018

  57. [65]

    Boone Loves Video. 2025. GeoLayers 3: Travel Map Animation Tutorial. https: //www.youtube.com/watch?v=qdrvtU6WQ0

    2025

  58. [66]

    Bryan Wang, Yuliang Li, Zhaoyang Lv, Haijun Xia, Yan Xu, and Raj Sodhi. 2024. LAVE: LLM-Powered Agent Assistance and Language Augmentation for Video Editing. In Proceedings of the 29th International Conference on Intelligent User Interfaces. 699–714

    2024

  59. [67]

    Haochen Li1 Jonathan Leung1 Hao Wang and Zhiqi Shen. [n.d.]. Prompting LLMs to Solve Complex Tasks: A. ([n. d.])

  60. [68]

    Lei Wang, Wanyu Xu, Yihuai Lan, Zhiqiang Hu, Yunshi Lan, Roy Ka-Wei Lee, and Ee-Peng Lim. 2023. Plan-and-solve prompting: Improving zero-shot chain- of-thought reasoning by large language models. arXiv preprint arXiv:2305.04091 (2023)

    2023 arXiv

  61. [69]

    Zhijie Wang, Yuheng Huang, Da Song, Lei Ma, and Tianyi Zhang. 2024. Promptcharm: Text-to-image generation through multi-modal prompting and refinement. In Proceedings of the 2024 CHI Conference on Human Factors in Com- puting Systems. 1–21

    2024

  62. [70]

    MotionArray Tutorials. 2025. Animate Maps Like The WSJ | After Effects Tutorial. https://www.youtube.com/watch?v=1Pf2EwiP1W4

    2025

  63. [71]

    Denis Wood and John Fels. 1992. The Power of Maps . Guilford Press

    1992

  64. [72]

    Haijun Xia. 2020. Crosspower: Bridging graphics and linguistics. In Proceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology . 722–734

    2020

  65. [73]

    Haijun Xia, Tony Wang, Aditya Gunturu, Peiling Jiang, William Duan, and Xiaoshuo Yao. 2023. CrossTalk: Intelligent Substrates for Language-Oriented Interaction in Video-Based Communication and Collaboration. In Proceedings of the 36th Annual ACM Symposium on User Interface Sof...

    2023

  66. [74]

    Zihan Yan, Chunxu Yang, Qihao Liang, and Xiang’Anthony’ Chen. 2023. XCre- ation: A graph-based crossmodal generative creativity support tool. InProceedings of the 36th Annual ACM Symposium on User Interface Software and Technology . 1–15

    2023

  67. [75]

    Hui Ye, Chufeng Xiao, Jiaye Leng, Pengfei Xu, and Hongbo Fu. 2025. Mo- GraphGPT: Creating Interactive Scenes Using Modular LLM and Graphical Con- trol. arXiv preprint arXiv:2502.04983 (2025)

    2025 arXiv

  68. [76]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems 35 (2022), 24824–24837

    2022

  69. [82]

    Lei Zhang and Steve Oney. 2020. Flowmatic: An immersive authoring tool for creating interactive scenes in virtual reality. In Proceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology . 342–353. LIST OF IMAGE SOURCES In Figure 2 and 4 we have used ...

    2020

  70. [2014]

    In Proceedings of the SIGCHI conference on human factors in computing systems

    Newsviews: an automated pipeline for creating custom geovisualizations for news. In Proceedings of the SIGCHI conference on human factors in computing systems. 3005–3014

  71. [2016]

    In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems

    Motion amplifiers: sketching dynamic illustrations using the principles of 2D animation. In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems. 4599–4609

    2016

  72. [2022]

    In Proceedings of the 35th annual ACM symposium on user interface software and technology

    Realitytalk: Real-time speech-driven augmented presentation for ar live storytelling. In Proceedings of the 35th annual ACM symposium on user interface software and technology. 1–12

  73. [2023]

    arXiv preprint arXiv:2306.00020 (2023)

    GPT4GEO: How a Language Model Sees the World’s Geography. arXiv preprint arXiv:2306.00020 (2023)

    2023 arXiv

  74. [2024]

    In Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology

    DrawTalking: Building Interactive Worlds by Sketching and Speaking. In Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology. 1–25

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.