Generative AI and world models

Publishing authors

Base scenario: 3.8k publishing authors. Paper count × 3.78 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,640 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 2,989 papers (33.5%). 15.0% of authorship records lack an Author ID; 14.2% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 3.78; 95% bootstrap interval 3.60–3.95. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $1.91B.

Creative, music, video content: $4.19B × 28.00%
Entertainment: $3.66B × 20.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

Scope and related research

Grouped by the output task; the underlying methods also apply to other modalities.

Organizations

Selected examples

OpenAI Models
Native image generation and editing in multimodal models.
Google DeepMind Models
Image generation and editing models in the Gemini model family.
Black Forest Labs Models
Image-generation models.

More organizations (9)

Adobe / Firefly Product
Firefly generates images from text prompts.
Ideogram Product
Text-to-image generation and editing workflows.
Kuaishou / Kling AI Product
AI video generation.
Lightricks Product
Generative creative tools.
Luma AI Product
Visual generative models.
Midjourney Models
Image and video generation models.
Photoroom Product
Product imagery and editing.
Recraft Product
Generative design.
Stability AI Models
Generative media models.

Video generation and editing

Text- and image-to-video, temporal consistency, scene control, and video editing.

Text-to-video Video editing

954 papers

3.4× vs. 2023

Authors · observed: 4,159
Investment · scenario: $1.04B–$4.15B

Sources and methodology

Publication activity

2019: 39
2020: 58
2021: 72
2022: 106
2023: 283
2024: 696
2025: 954

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("text to video" OR "video generation" OR "video diffusion" OR "video editing") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence") OR "diffusion model" OR "generative model")

Retrieved 2026-09-26.

Publishing authors

4,159 distinct OpenAlex Author IDs across all 954 matching papers. This is an observed count within the query result; no publication-rate assumption is used.

Full query result: 954 of 954 papers (100.0%). 15.9% of authorship records lack an Author ID; 10.2% of retrieved works lack an abstract. Retrieved 2026-09-26.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $2.07B.

Creative, music, video content: $4.19B × 32.00%
Entertainment: $3.66B × 20.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

Scope and related research

Action-conditioned interactive models also appear under world models.

Organizations

Selected examples

Google DeepMind Models
Veo generates video with synchronized audio.
Runway Models
Video and world models.
Luma AI Product
Visual generative models.

More organizations (10)

Adobe / Firefly Product
Adobe Firefly generates video from text prompts and reference images.
ByteDance Seed Models
Multimodal model research.
Decart Models
Real-time generative video.
HeyGen Product
Talking avatars and video translation.
Kuaishou / Kling AI Product
AI video generation.
Lightricks Product
Generative creative tools.
Midjourney Models
Image and video generation models.
Mirage / Captions Product
AI video models, avatars and editing tools.
Pika Product
AI video generation.
Synthesia Product
AI avatars and business video.

3D generation and neural rendering

Text- and image-to-3D, scene representations, NeRF, Gaussian splatting, and novel-view synthesis.

3D generation Gaussian splatting NeRF

2,641 papers

2.3× vs. 2023

Authors · scenario: 4.2k–12k
Investment · scenario: $392M–$1.57B

Sources and methodology

Publication activity

2019: 43
2020: 85
2021: 237
2022: 455
2023: 1,148
2024: 2,251
2025: 2,641

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("neural radiance field" OR "Gaussian splatting" OR "text to 3D" OR "3D generation" OR "neural rendering" OR "novel view synthesis")

Retrieved 2026-09-26.

Publishing authors

Base scenario: 4.2k publishing authors. Paper count × 4.68 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 4,155 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 2,641 papers (37.9%). 13.7% of authorship records lack an Author ID; 13.3% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 4.68; 95% bootstrap interval 4.45–4.95. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $785M.

Creative, music, video content: $4.19B × 10.00%
Entertainment: $3.66B × 10.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

Scope and related research

Overlaps with geometric vision, especially when reconstructing and rendering the same scene.

Organizations

Selected examples

Microsoft Research
TRELLIS generates 3D assets from text and images.
Meshy Product
Generative 3D.
Tripo AI Product
Generative 3D.

More organizations (5)

Kaedim Product
AI-assisted 3D assets.
Sloyd Product
3D asset generation.
SpAItial Models
Spatial foundation models.
Tencent / Hunyuan Models
Hunyuan3D generates textured 3D assets and provides model weights and code.
World Labs Models
Spatial intelligence and world models.

Motion generation and avatars

Text-to-motion, speech-driven gestures, talking heads, retargeting, and digital characters.

Digital avatars Text-to-motion

553 papers

1.8× vs. 2023

Authors · observed: 2,132
Investment · scenario: $576M–$2.30B

Sources and methodology

Publication activity

2019: 105
2020: 103
2021: 125
2022: 174
2023: 301
2024: 463
2025: 553

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("motion generation" OR "text to motion" OR "talking head generation" OR "gesture generation" OR "motion retargeting" OR "animatable avatar")

Retrieved 2026-09-26.

Publishing authors

2,132 distinct OpenAlex Author IDs across all 553 matching papers. This is an observed count within the query result; no publication-rate assumption is used.

Full query result: 553 of 553 papers (100.0%). 14.3% of authorship records lack an Author ID; 11.6% of retrieved works lack an abstract. Retrieved 2026-09-26.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $1.15B.

Creative, music, video content: $4.19B × 10.00%
Entertainment: $3.66B × 20.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

Scope and related research

Creates movement and appearance. Pose estimation and human recognition are covered under vision.

Organizations

Selected examples

DeepMotion Product
SayMotion generates 3D character animation from text.
Motorica Product
Generative character animation.
Kinetix Product
Motion and character animation.

More organizations (8)

Autodesk / Wonder Dynamics Product
AI-assisted character animation and visual-effects workflows.
Part of Autodesk
Brahma / Metaphysic Product
AI-assisted digital-human content creation.
Part of Brahma / DNEG Group
D-ID Product
Digital avatars.
HeyGen Product
Talking avatars and video translation.
Mirage / Captions Product
AI video models, avatars and editing tools.
Move AI Product
Markerless motion capture.
Plask Product
AI motion capture.
Synthesia Product
AI avatars and business video.

World models and simulation

Learned dynamics, action-conditioned generation, interactive environments, and simulation.

World models Action-conditioned simulation

510 papers

3.4× vs. 2023

Authors · observed: 1,556
Investment · scenario: $1.41B–$5.63B

Sources and methodology

Publication activity

2019: 23
2020: 49
2021: 77
2022: 83
2023: 150
2024: 231
2025: 510

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("world model" OR "action conditioned generation" OR "interactive simulation") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence") OR "reinforcement learning")

Retrieved 2026-09-26.

Publishing authors

1,556 distinct OpenAlex Author IDs across all 510 matching papers. This is an observed count within the query result; no publication-rate assumption is used.

Full query result: 510 of 510 papers (100.0%). 17.2% of authorship records lack an Author ID; 10.2% of retrieved works lack an abstract. Retrieved 2026-09-26.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $2.81B.

AI infrastructure/models/research/governance: $143.2B × 1.00%
Robotics: $7.84B × 10.00%
Energy management: $4.64B × 5.00%
Entertainment: $3.66B × 10.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.