The AI Moodboard Brief Checklist (2026): What to Write Before You Generate Anything

Article Metadata

By Justkay, Documentary Filmmaker and Founder of Storyflow Published September 2, 2026 · 15 min read · Moodboards

Breadcrumb

Home > Blog > Moodboards > AI Moodboard Brief Checklist 2026

Table of Contents

1. Why the Brief Decides the Board
2. The Six-Part Brief
3. The Negative Constraint
4. The Fifteen-Word Test
5. Generated vs Reference Moodboards
6. The Licensing Question
7. 8 AI Moodboard Tools
8. How We Evaluated These Tools
9. Tool Reviews
10. The Captioning Step Everyone Skips
11. FAQ: AI Moodboard Briefs
12. The Bottom Line
13. Author
14. Related Reading

Key Takeaways

• Six constraints, not a subject. Subject-only prompts return the average of the training data.
• The negative constraint does more work than the other five combined.
• If your brief fits in fifteen words, it is a search query and the output will be generic.
• Generated boards show a feeling. Reference boards show what is achievable. They are different documents.
• An uncaptioned moodboard transmits a mood and no instruction, which is why clients approve them and then reject the work.

Why the Brief Decides the Board

Every AI image generator will return something competent from three words, and that is precisely the failure mode. A competent generic image is harder to reject than an obviously bad one, so subject-only prompts produce boards that look fine and communicate nothing.

The mechanism is simple. A generator asked for "moody coffee shop" returns the centre of mass of everything it has seen labelled that way: warm tungsten, a wooden counter, steam, shallow depth of field. That image is not wrong. It is the same image everyone else gets, and a moodboard's entire job is to say what is specific about this project.

Constraints move you off the average. Each one you add narrows the space the generator is sampling from, and the narrowing is the value.

The Six-Part Brief

Write all six before generating anything. Ten minutes, and it changes the output more than switching tools does.

1. Subject. What is literally in frame. Be concrete: not "a workspace" but "a desk at 6am with one lamp on".

2. Emotional register. What the viewer should feel. This is the part most briefs skip and the part that separates two boards with identical subjects. "Calm and slightly lonely" and "calm and safe" produce very different images of the same desk.

3. Colour and light. Where the light comes from, its quality, its temperature, and the palette. "Single hard source from frame left, cool daylight, desaturated except one warm accent" is a brief. "Nice lighting" is not.

4. Texture and materials. What things are made of. Concrete, linen, brushed steel, worn paper. Materials carry more of a look than people expect and they are the easiest thing to specify precisely.

5. Composition. Framing, distance, lens feel, where the subject sits. "Wide, low camera, subject small in frame, lots of negative space above" is directable.

6. What it must not become. The negative constraint, below.

A brief with all six is typically forty to eighty words. That length is not verbosity; it is the difference between sampling a broad region and a narrow one.

The Negative Constraint

The single highest-leverage line in an AI moodboard brief, and almost nobody writes it.

Generators drift toward the average, and the average has a recognisable character per subject: corporate stock for offices, warm cinematic teal for anything moody, glossy product photography for objects, and smiling people for anything with a person in it. If you do not say what to avoid, you get that character, because it is where the mass of the training data sits.

Useful negative constraints are specific:

• "Not stock photography. No smiling models, no clean desks, no laptops with generic dashboards."
• "Not cinematic teal and orange. No lens flare."
• "Not glossy. Nothing looks new."
• "Not a magazine spread. This should look like something someone took, not something someone art-directed."
• "No aerial shots, no drone perspective."

Write two or three, and make them about what your subject specifically drifts toward. A generic "not generic" does nothing. Naming the exact cliché your subject attracts does most of the work.

The Fifteen-Word Test

A quick check before you generate: read your brief and count the words.

Under fifteen words, you have written a search query, not a brief, and you will get search-query output: the average, competently rendered. This is not a rule about verbosity for its own sake; it is that six constraints cannot fit in fifteen words.

Forty to eighty words with all six parts present is where output stops being generic. Beyond about 150 words most generators start ignoring parts of the prompt, so more is not linearly better.

The other half of the test: could a photographer shoot from your brief? If a person could not execute it, a generator is guessing at the same gaps, and it will fill them with the average.

Generated vs Reference Moodboards

Worth separating, because they are different documents and confusing them causes real problems.

A generated moodboard shows a feeling that does not exist yet. It is fast, it is exactly on brief, and it is free of the constraint of reality. Its danger is that it promises something nobody has proved is achievable: generated images have no budget, no location, no talent and no physics.

A reference moodboard is real images of real work. It is slower to gather, less exactly on brief, and it proves the thing is possible. When a client approves a reference board, they have approved something someone actually made.

The failure this causes: a client approves a generated board and expects the shoot to look like it. Generated light is often physically impossible, generated architecture does not exist, and generated faces are nobody you can cast.

The practical answer is both, labelled:

• Generated images for the feeling and direction, marked as such.
• Reference images for anything the shoot must actually reproduce: a lighting setup, a location type, a wardrobe, a camera move.

Saying which is which on the board itself prevents the entire class of problem.

The Licensing Question

Worth checking before commercial work, and the answer differs by tool.

Generators differ in what they say about commercial use, indemnification and the provenance of training data, and those terms change. Adobe Firefly markets itself on being trained on licensed and public-domain content with an indemnification offer for enterprise customers, which is why it appears on brand work where legal has an opinion. Others are less explicit.

Two practical rules that hold regardless of vendor:

An internal moodboard is a much lower-risk use than a delivered asset. A board that shapes a shoot is not a published image, and most concerns are about the latter.

Check the current terms yourself before commercial delivery, because they change and no article stays accurate. If a generated image is going out in a campaign, that is a legal question rather than a design one.

8 AI Moodboard Tools

Prices are publicly listed plans as of early 2026 and change often. Ratings weigh image quality, how easily images become a board someone can read, licensing clarity for commercial work, and cost.

How We Evaluated These Tools

Four criteria:

1. Output quality against a six-part brief, rather than against a three-word prompt, since that is how these tools should be used.
2. Board assembly. A folder of images is not a moodboard; getting from generation to a captioned, shareable board is most of the work.
3. Licensing clarity for commercial use.
4. Cost against a realistic monthly image count, which is higher than people estimate because iterating on a brief costs generations.

Tool Reviews

1. Midjourney

The best individual images in this category and still improving faster than most. Given a six-part brief it responds to every part of it, which is not true of every generator, and its style control once learned is the deepest here.

It has no board. You generate, download and assemble somewhere else, which is a real workflow cost when a moodboard is twenty images. Its licensing terms are less explicit than Adobe's for commercial delivery.

Pricing: from around $10 per month.

Pros: the strongest image quality; responds well to detailed briefs; deep style control.

Cons: no board assembly at all; prompting skill required; check terms before commercial delivery.

2. Canva

The fastest route from an idea to a board somebody can open. Generation, layout, captioning and sharing in one place, and it is already paid for in most teams. For a moodboard whose job is to be sent, that integration beats better images in a folder.

Image quality is a step behind Midjourney and Firefly, and its layouts can flatten a distinctive look into a template.

Pricing: Free tier; Pro around $15 per month.

Pros: generation and board assembly in one tool; fastest to something shareable; already in most teams.

Cons: images a step behind the specialists; template look; limited style control.

3. Adobe Firefly

The choice when commercial licensing matters. Adobe's position on training data and its indemnification offer for enterprise customers is why Firefly appears on brand work where legal is involved, and its integration with the rest of Creative Cloud makes the path from generation to layout short.

Style control is narrower than Midjourney's and image quality trails it slightly.

Pricing: included with Creative Cloud plans; standalone from around $5 per month.

Pros: the clearest commercial licensing story here; strong Creative Cloud integration; good quality.

Cons: narrower style range than Midjourney; another Adobe subscription if you are not already in it.

4. Storyflow

Storyflow is our product and it is fourth, behind Midjourney, Canva and Firefly, which is the honest placement. Its image quality is good rather than best, and anyone whose priority is the strongest possible individual frame should use Midjourney and assemble elsewhere. AI image generation also requires Pro at $14 per month annually or above, not the entry Plus tier, which matters for this use case specifically.

Its case is narrower and real: generation happens onto a board that already holds the brief, so the six constraints sit next to the images they produced, and the AI reads the whole board before generating rather than working from a prompt box in isolation. On a project where the moodboard is one part of a larger plan, that removes the folder-of-downloads step entirely.

Storyflow is paid-only during early access; its Free plan launches before the end of 2026, and anyone a paid member invites to a board joins free now.

Pricing: Plus $7.99 per month billed annually ($9.99 monthly); Pro $14 ($19 monthly) adds AI image generation; Max $39 ($49 monthly).

Pros: brief and images on the same board; whole-board AI context; flat per-account pricing; no download-and-assemble step.

Cons: image quality behind Midjourney and Firefly; AI images need Pro not Plus; less licensing clarity than Adobe for commercial delivery.

5. Krea

The best style consistency here, which matters if a board needs twelve images that feel like one shoot rather than twelve separate ideas. Real control over style references.

Weaker board assembly and a genuine learning curve.

Pricing: from around $10 per month.

Pros: the best consistency across a set; strong style referencing.

Cons: assembly happens elsewhere; learning curve; smaller ecosystem.

6. Milanote

No generation at all, and the best board in this list. For a reference moodboard, which is the document that proves a look is achievable, Milanote is the strongest option here and the calmest to read.

If your requirement is generation, this is not the tool.

Pricing: Free for 100 items; Pro $9.99 per month.

Pros: the best reference board; excellent image handling; calm to read.

Cons: no generation; the free tier's 100-item cap is reached by one moodboard.

7. Pinterest

Still the fastest way to gather real reference, and reference is half of the answer. Free, enormous, and organised by exactly the visual associations you are looking for.

No generation, weak as a client-facing deliverable, and its recommendations pull you toward the popular, which is another route to the average.

Pricing: free.

Pros: the fastest reference gathering; free; vast.

Cons: no generation; not a deliverable; recommendation drift toward the generic.

8. Figma with plugins

Sensible if the team is already in Figma. Generation via plugins is variable and layout is excellent. The plugin dependency means quality and availability move without warning.

Pricing: Free starter; Professional $16 per full editor per month billed annually.

Pros: excellent layout; already there for design teams.

Cons: generation quality depends on third-party plugins; highest editor cost.

The Captioning Step Everyone Skips

A moodboard without captions transmits a mood and no instruction. This is the single most common defect in moodboards generated or gathered, and it is why clients approve boards and then reject the work.

Caption every image with what it is for and what it is not:

• "This is the light, not the colour."
• "This is the framing. Ignore the wardrobe."
• "This is the texture we want on the walls."
• "This is the closest reference for the mood, but our version is warmer."

Six captioned images beat forty uncaptioned ones. An uncaptioned board lets the client approve their own interpretation, and their interpretation and yours diverge in exactly the places nobody discussed.

A useful test: hand the board to somebody who was not in the conversation and ask what they think the shoot will look like. Whatever they say wrong is a missing caption.

The Bottom Line

Write the six-part brief before you open any generator, and spend the most effort on the line that says what the image must not become. Count the words: under fifteen and you have written a search query. Then caption every image with what it is for and what to ignore, and label which images are generated direction and which are real reference the shoot has to reproduce. The tool matters far less than any of that.

Author

Justkay is a documentary filmmaker and the founder of Storyflow. The generated-versus-reference section exists because he once had a lighting setup approved that was not physically possible.