World building
For fantasy and sci-fi movies, AI could quickly generate backdrops and action instead of requiring some real sets and props. [1]
AI × CREATIVITY × THE FUTURE OF FILM
The poster is only
the first frame.
Step inside Synthesising Hollywood. Direct a scene, follow the models, and look behind the claims.
No app or account needed. Start here, then point your camera at the whole printed poster.
01 / A REAL MODEL OUTPUT
From ComfyUI’s official Wan 2.2 example. See its source ↗
A published demo, not a film made by our group.
CHOOSE YOUR NEXT FRAME
One film. Many decisions.
YOUR TURN
Make the creative choices.
Change the brief and see the storyboard respond. This is a sketch and prompt builder—not an AI video generator.
Your choices → a shot brief → human review
A useful prompt describes subject, action, setting and camera. The result still needs selection and revision. ComfyUI example [3] ↗
01 / INSIDE THE MODELS
Different machinery. Connected outputs.
These are four representative model families, not four steps that every system must follow.
02 / FROM IDEA TO RELEASE
Open the real ComfyUI workflow and its source ↗A human decision at every cut.
Tap a frame to see what goes in, what comes out, and what still needs your judgement.
03 / THE CREATIVE COST
Would you release it?
Try five production decisions. The strongest workflow includes consent, review and responsibility.
Our group’s pros & cons
Consistent style & props
It is hard to keep props and style consistent throughout a movie. With AI videos, referencing images can help, though results still need checking. [1,3]
Low cost for redos
AI-generated shots can be cheaper and easier to regenerate than traditional reshoots, which may require rebuilding sets and recalling crews. [1]
Teaching & demonstrations
Supports teaching and topic demonstrations (e.g., Google NotebookLM), which could produce teaching videos and slide decks. [26]
Creative ideas
Lets people imagine themselves or their IPs in different situations, making prompts a strong source of creative ideas. [1]
Low variety
The base model could produce very similar scenes or characters. Copying others’ work or using someone’s likeness without permission can raise rights concerns. [5,20]
Deepfake abuse
Shorts could feature real people in undesirable or sexual settings, harming their reputations. Videos of loved ones may be used to threaten or scam family members. [19,23]
Labour & consent
AI could affect traditional film workers’ jobs, wages and creative control. Reusing someone’s appearance or voice raises consent and credit concerns. [6,20]
Reality or fiction?
As AI videos become more realistic, people may become confused about what is real and more vulnerable to fake news. [21]
Revenue limits
YouTube may reject repetitively produced AI junk, while people making similar videos may also face monetisation restrictions. [10]
Numbers refer to the full bibliography. These are the captions from our poster.
Turn the poster into a doorway.
Open the full-screen camera and frame the four corner markers. When the poster is recognised, video and interactive hotspots follow its position and perspective.
Keep at least three corner markers visible in good light. The overlay hides if tracking is lost.