How to Use a Mood Board for an AI Film Without Copying It

Why models copy the board

An image model given a reference picture treats it as evidence of what you want, and the strongest evidence in a photograph is its content: the subject, the pose, the composition, even a watermark in the corner. Ask for "the mood of this image" and the result can come back as the image itself with small changes. In video it goes further, because an image-to-video model starts from whatever frame you give it.

That is a craft problem before it is a legal one. A film built from near-copies of other people's frames looks like a collage of its references instead of a world of its own.

What the law protects, and what it leaves open

In the US, style is free and expression is not. The Copyright Act says protection never extends "to any idea," and the US Copyright Office's July 2024 report on digital replicas says copyright "does not protect artistic style as a separate element of a work." The same report adds that the law "may, however, provide a remedy" when an output in someone's style "ends up replicating not just the artist's style but protectible elements of a particular work." That is the line a mood board has to stay behind. The feel of a photographer's light is yours to use, and recreating one of their photographs, its particular composition and subject, is where the risk starts.

The Copyright Office's January 2025 report on copyrightability matters for what you can own. It says "prompts alone do not provide sufficient human control to make users of an AI system the authors of the output," but "where a human inputs their own copyrightable work and that work is perceptible in the output, they will be the author" of that portion. Using your own photographs and frames as references protects you twice. No one else's rights are in the frame, and your own expression is.

In the UK, Getty Images v Stability AI, decided on November 4, 2025, was narrow. The court dismissed Getty's secondary infringement claim because the model "does not store or reproduce any Copyright Works," and Getty won only narrow trademark points over old watermarks, findings the judge called "both historic and extremely limited in scope." Getty had dropped its training and output claims before judgment. In the US, the artists' case Andersen v. Stability AI, in which Runway is also a defendant, has a jury trial set for September 20, 2027. The law on reference images is still moving, so build boards you would be comfortable defending.

What contest rules ask for

Contest rules put the burden on you, and several go further than the law. The Runway AI Festival makes entrants "solely responsible for obtaining" all "required rights, releases, consents, permissions, clearances, licenses" and asks for "a short, written description of the AI techniques used." Curious Refuge's 2026 horror rules say "Your idea must be original and not connected to existing 3rd party Intellectual Property." Higgsfield disqualifies "characters, franchises, brands, logos, music, or other third-party IP used without authorization," and after its deadline it makes each project's "prompts, generation history, and all assets kept inside the project" public, so your references may be seen. Reply's rules require teams to buy licenses for images and music and to submit a PDF describing their use of generative AI. Golden Dunes Dubai says "The submitter must hold full rights to the film, including music, visuals and any AI-generated content."

The tools ask the same of you. Midjourney's terms require that "you have all necessary rights and permissions to provide the Content" you upload, and uploading grants Midjourney a license to reproduce it and prepare derivative works of it, a license you cannot give for someone else's photograph. Runway's terms say you must have "all rights, licenses, and permissions needed to provide Your Content," and Google's generative AI policy bars content that "Violates the rights of others, including privacy and intellectual property rights." Saving an image to a Pinterest board gives you no rights in it.

Turn the board into words

Look at the board and write one line for each thing it tells you. Most boards say five things.

• Light: Where it comes from, how hard it is, what time of day it feels like.
• Palette: Four to six colors as names plus hex codes, sampled with an eyedropper from the board.
• Texture: Grain, softness, halation, the surface of things.
• Energy: Still or restless, the pace of the edit, how close the camera sits.
• Era and world: The decade, the places, the wardrobe logic.

Then write the ban list, the specific things from the board that must not appear: a recognizable subject, a person's face, a location that belongs to one photograph, a logo, lettering, a signature composition. Those lines go into every prompt, and the board itself stays out of the model.

A board, translated

Here is the translation for a hypothetical board of eight images for a spec spot about a made-up coffee brand. The images stay on the wall, and this goes into the prompts.

Light: Low winter sun from frame left, hard, long shadows across the tables, warm practical lamps inside.
Palette: Espresso #2B1B14, oat #E8DCC8, brass #B08D57, slate blue #4A5A6A, cream #F6F1E7.
Texture: Fine 35mm grain, soft highlights with a little halation, no digital sharpening.
Energy: Unhurried. Locked-off frames or a slow dolly, people moving, camera still.
Era and world: Present day, a small city cafe, worn wood, brushed steel, wool coats.
Never: The corner cafe from image 3, the barista's face from image 5, any shop signage or lettering, the overhead cup composition from image 7.

The ban list names specific images from the board, because those are the pictures a model would reproduce if you fed them in. Everything above it describes the light and the world, which belong to no single photograph.

Use style references with a light touch

Style reference features exist for this job, and they still need a light hand. Midjourney says its Style Reference "doesn't copy objects or people, just the overall style," and you can turn it down with a style weight below the default of 100. A style reference built from your own frames, your own photographs or licensed images keeps the feeling and leaves the content. The kind of reference to avoid is someone else's photograph used as an image prompt or as the first frame of an image-to-video shot, because a video model treats that picture as the start of the shot.

Build the board from sources you can defend

The safest board is one you can show a festival or a client without flinching. Use your own photographs and frames, stills from your own earlier work, public-domain images, and stock you have licensed for reference use. Keep a simple log with each image's source and license, and note which ones went into a model as references and which stayed on the wall. Some festivals ask how a film was made, and the We Are Human Festival requires an Ethics Notebook with every entry; a reference log answers those questions in a minute.

How Overs reads a mood board

Overs treats a mood board the way this page recommends. You add up to twelve images and a note, and it reads the feeling of the board, the light, the mood and the energy, then applies that to your product in your brand's style instead of copying what is in the pictures. A dial decides whether the brand's own look or the board leads. Turn it toward the brand for a client who guards their look, and toward the board for a one-off campaign with its own world.

Overs also studies the brand from its website, its photos and an optional guidelines PDF, writes the visual DNA with hex colors, light, lens and grain, and applies brand rules last. You see the shot plan and an estimated cost before anything renders, and you approve, change or reject every photo.

The process comes from 100 Creatives, the agency Abhi Chawla founded, which used it to make hundreds of photos a week for hundreds of brands. Across four of its 2026 campaigns, 361 renders gave 45 keepers, about 8 tries per usable photo, which is $0.27 to $1.69 in AI fees per keeper at published prices of $0.034 to $0.211 a render. Abhi's own businesses make their images with Overs today. It makes stills, and you animate them in your video tool, starting from the motion prompt Overs writes for each photo for Seedance or Veo. Overs' guide to using a mood board with AI without copying it has a template for turning a board into words.

Overs is a sister product to AI Film Contests, from the same founder. The free plan covers 40 photos a month on your own OpenRouter key, with no markup from Overs. Try Overs free at www.overs.studio and give it the board for your next spot.