Workflow 1
ComfyUI official Wan 2.2 positive-prompt node screenshot. Cropped from the official workflow screenshot.
Image source [3] ↓Current poster image credits
Image credits identify where a visual comes from. A caption’s numbered citation supports its claim; these serve different purposes.
Workflow 2 · Image-model example
Comfy-Org armoured-character input image for Wan 2.2 14B I2V. This is a published model input, not a photograph or output from the illustrated U-Net.
Image source [3] ↓Workflow 3
ComfyUI KSampler and Save Video: two crops from the official workflow screenshot.
Image source [3] ↓Video-model example
Three frames from the Comfy-Org Wan 2.2 14B I2V demonstration, at 0.5, 2.0 and 3.5 seconds. Arrows indicate playback order.
Image source [3] ↓Workflow 4
Blackmagic Design DaVinci Resolve editing-interface image. Official promotional interface image.
Image source [13] ↓Workflow 5
Native Foreign / Toys“R”Us AI-assisted brand-film train image. Published production-case image.
Image source [17] ↓Workflow 6
Waymark / Latent Cinema, The Frost film-player webpage. Cropped webpage screenshot.
Image source [11] ↓Direct Comfy-Org assets: input image · example clip.
These visuals show separate published examples, not one film produced by our group. AI-generated film imagery is not documentary photography.
Original diagrams & decorative artwork
The four model diagrams are original explanatory SVG drawings prepared with AI assistance, based on these sources. They are not reproduced figures or measured model activations.
AI-generated decorative icons
The ten pros/cons icons were generated using OpenAI’s image-generation tool for this poster. They illustrate concepts and are not source evidence, photographs or figures from the cited papers. No model version is asserted where the generation record does not name one.
The film strip, clapperboard and paper texture are decorative HTML/SVG artwork.
Numbered references
Numbers 1–26 are preserved so existing links remain stable. The current poster cites 23 of these entries. References [12] and [15] belong to earlier examples; [16] is retained disclosure background. They are labelled below.
- 1
Zhang, R., Yu, B., Min, J., Xin, Y., Wei, Z., Shi, J. N., Huang, M., Kong, X., Xin, N. L., Jiang, S., Bahuguna, P., Chan, M., Hora, K., Yang, L., Liang, Y., Bian, R., Liu, Y., Valencia, I. C., Tredinick, P. M., … Rao, A. (2025). Generative AI for film creation: A survey of recent advances. arXiv. https://doi.org/10.48550/arXiv.2504.08296
Overview of AI film production, visual styles and practical limitations.
Open source ↗https://arxiv.org/abs/2504.08296
- 2
Rombach, R., Blattmann, A., Lorenz, D., Esser, P., & Ommer, B. (2022). High-resolution image synthesis with latent diffusion models. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10684–10695.
Image-model diagram: latent diffusion, conditioning and decoding.
Open source ↗https://openaccess.thecvf.com/content/CVPR2022/html/Rombach_High-Resolution_Image_Synthesis_With_Latent_Diffusion_Models_CVPR_2022_paper.html
- 3
ComfyUI. (n.d.). Wan2.2 video generation: ComfyUI official native workflow example. ComfyUI documentation. Retrieved November 7, 2026.
Official workflow, node screenshots and separate image-to-video reference input.
Open source ↗https://docs.comfy.org/tutorials/video/wan/wan2_2
- 4
Copet, J., Kreuk, F., Gat, I., Remez, T., Kant, D., Synnaeve, G., Adi, Y., & Défossez, A. (2023). Simple and controllable music generation. Advances in Neural Information Processing Systems, 36. Paper and audio demonstrations.
MusicGen audio-token architecture and text- or melody-conditioned music examples.
Open source ↗https://ai.honu.io/papers/musicgen/
- 5
Doshi, A. R., & Hauser, O. P. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. Science Advances, 10(28), eadn5290. https://doi.org/10.1126/sciadv.adn5290
Evidence about short-story ratings and similarity. The poster does not treat this as a measured result for films.
Open source ↗https://doi.org/10.1126/sciadv.adn5290
- 6
Halperin, B. A., & Rosner, D. K. (2025). “AI is soulless”: Hollywood film workers’ strike and emerging perceptions of generative cinema. ACM Transactions on Computer-Human Interaction, 32(2), 1–27. https://doi.org/10.1145/3716135
Film workers’ perspectives on AI, creative labour and control.
Open source ↗https://doi.org/10.1145/3716135
- 7
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30.
Transformer components: multi-head attention, residual paths and feed-forward layers.
Open source ↗https://arxiv.org/abs/1706.03762
- 8
Hugging Face. (n.d.). Causal language modeling. Transformers documentation. Retrieved November 7, 2026.
Autoregressive language modelling and next-token prediction.
Open source ↗https://huggingface.co/docs/transformers/main/en/tasks/language_modeling
- 9
Team Wan. (2025). Wan: Open and advanced large-scale video generative models. arXiv. https://doi.org/10.48550/arXiv.2503.20314
Video diffusion Transformer, spatiotemporal modelling and video VAE. See the source for the full contributor list.
Open source ↗https://arxiv.org/abs/2503.20314
- 10
Google. (n.d.). YouTube channel monetization policies. YouTube Help. Retrieved November 7, 2026.
Platform revenue eligibility and restrictions on repetitive or mass-produced content.
Open source ↗https://support.google.com/youtube/answer/1311392
- 11
Waymark & Latent Cinema. (n.d.). The Frost. Official film website. Retrieved November 7, 2026.
Published AI-film example, synopsis, credits and online player.
Open source ↗https://www.thefrost.ai/
- 12
Earlier-version example; not a current poster image.
Runway. (2024). AI Film Festival 2024. Festival programme: A Tree Once Grew Here, by John and Dara Semerad.
Published film example for discussion of visual styles.
Open source ↗https://aif.runwayml.com/2024
- 13
Blackmagic Design. (n.d.). DaVinci Resolve: Edit. Official product documentation. Retrieved November 7, 2026.
Real editing-interface image used in the sound and editing stage.
Open source ↗https://www.blackmagicdesign.com/products/davinciresolve/edit
- 14
Meta FAIR. (n.d.). MusicGen model card. AudioCraft documentation. Retrieved November 7, 2026.
Model description and the CC BY-NC 4.0 licence for model weights.
Open source ↗https://facebookresearch.github.io/audiocraft/model_cards/MUSICGEN_MODEL_CARD.html
- 15
Earlier-version example; not a current poster image.
SAG-AFTRA. (2023). L.A. takes to the picket lines. SAG-AFTRA.
Primary industry source for the 2023 actors’ strike.
Open source ↗https://www.sagaftra.org/la-takes-picket-lines
- 16
Retained background on disclosure; not cited in the current poster.
Google. (n.d.). Disclosing use of altered or synthetic content. YouTube Help. Retrieved November 7, 2026.
Disclosure requirements for realistic altered or synthetic content.
Open source ↗https://support.google.com/youtube/answer/14328491
- 17
Native Foreign. (2024). Toys“R”Us AI film. Official production case study.
Real AI-assisted brand-film example for the commercial-production stage; no earnings are claimed.
Open source ↗https://nativeforeign.co/work/toys-r-us-ai-film
- 18
Writers Guild of America. (2023). Summary of the 2023 WGA MBA. 2023 contract campaign.
Primary industry source for the agreement’s AI provisions.
Open source ↗https://www.wgacontract2023.org/the-campaign/summary-of-the-2023-wga-mba
- 19
eSafety Commissioner. (n.d.). Image-based abuse. Australian Government. Retrieved November 7, 2026.
Non-consensual intimate images, including AI deepfakes, and harm to the people depicted.
Open source ↗https://www.esafety.gov.au/key-topics/image-based-abuse
- 20
U.S. Copyright Office. (2024). Copyright and artificial intelligence, Part 1: Digital replicas. U.S. Copyright Office.
Risks involving unauthorised likenesses; distinct from low variety and not a claim that all similar outputs infringe rights.
Open source ↗https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf
- 21
Vaccari, C., & Chadwick, A. (2020). Deepfakes and disinformation: Exploring the impact of synthetic political video on deception, uncertainty, and trust in news. Social Media + Society, 6(1). https://doi.org/10.1177/2056305120903408
Experimental evidence about uncertainty and reduced trust in news after exposure to political deepfakes; not evidence that all AI films deceive viewers.
Open source ↗https://doi.org/10.1177/2056305120903408
- 22
SAG-AFTRA. (2026, February 13). SAG-AFTRA statement on Seedance 2.0. SAG-AFTRA.
2026 criticism of unauthorised voices and likenesses; separate from the 2023 strikes.
Open source ↗https://www.sagaftra.org/sag-aftra-statement-seedance-20
- 23
Financial Crimes Enforcement Network. (2024, November 13). FinCEN alert on fraud schemes involving deepfake media targeting financial institutions. FIN-2024-Alert004. U.S. Department of the Treasury.
Deepfake voices or videos used to impersonate family members and friends in family-emergency scams.
Open source ↗https://www.fincen.gov/system/files/shared/FinCEN-Alert-DeepFakes-Alert508FINAL.pdf
- 24
Apple. (n.d.). Final Cut Pro. Apple. Retrieved November 7, 2026.
Official video-editing product information.
Open source ↗https://www.apple.com/final-cut-pro/
- 25
Apple. (n.d.). Logic Pro. Apple. Retrieved November 7, 2026.
Official audio production and editing product information.
Open source ↗https://www.apple.com/logic-pro/
- 26
Google Workspace. (2026, March). New ways to customize and interact with your content in NotebookLM. Google Workspace Updates.
AI-generated video overviews and slide decks for teaching and demonstrations.
Open source ↗https://workspaceupdates.googleblog.com/2026/03/new-ways-to-customize-and-interact-with-your-content-in-NotebookLM.html