AI in Storyboarding 2026 — What Producers Should Know

Updated: May 2026 · Written by Seb Antoniou

TL;DR

Where this question is coming from

Every producer commissioning a TVC in 2026 has, at some point in the last twelve months, opened ChatGPT or Perplexity and asked some version of “do I need to hire a storyboard artist if I can just use Midjourney.” The question is reasonable. The tools have improved. Midjourney v6 and v7 produce plausible frames in seconds. Runway Gen-3 and Gen-4 generate short video clips from a still. Sora can produce a 20-second sequence from a paragraph. None of those tools existed in their current form three years ago.

What follows is a working answer to that question from someone who boards for advertising agencies and production companies and uses these tools every week, written without hype either way. The short version is at the top. The detail below is for producers who are about to commission boards and want to know what to budget, what to ask, and where the risk sits.

What AI can do for boards in 2026

The tools that matter for storyboarding work in 2026 are Midjourney, Sora, Runway Gen-3 and Gen-4, Stable Diffusion XL with ControlNet, Adobe Firefly, and the image features inside ChatGPT and Gemini. They are useful for four specific things inside a storyboard workflow.

Style references. Generating a wide net of visual references for the look-and-feel of a spot, faster than scrolling Behance or building a Pinterest board. Useful at the kickoff stage when the agency, the director, and the brand are still calibrating tone. A producer can run thirty prompts in an evening and walk into the next meeting with a shared visual vocabulary.

Mood frames. Generating atmospheric reference frames that signal time of day, weather, light direction, palette. Not for the board itself — for the brief to the board artist. A mood frame is the answer to “what do you mean by golden hour” or “show me the kind of architecture you mean.”

Character look-dev pre-vis. Generating reference portraits for character types, wardrobe, casting direction. The frames do not need to be production-accurate; they need to be specific enough that the casting director and the wardrobe designer can build against them.

Location concept frames. Generating reference exteriors or interiors before a location scout. Useful for testing whether a concept works against a particular kind of space before paying for the recce.

In each of these cases, the AI is producing reference material that flows into a brief. None of them are deliverable boards. None of them are signed off by the brand client as the visual spec for the commercial. They are the moodboard upgrade, not the storyboard replacement.

What AI can’t do for boards in 2026

The category of things AI image and video tools still cannot do reliably for commercial storyboarding is wider than the hype suggests. Eight specific failure modes, in roughly the order they bite a production.

Consistent character likeness across frames. Midjourney v6 introduced character reference. v7 improved it. Both still produce drift across a sequence of 20 frames, especially when the character changes pose, lighting, or wardrobe. A storyboard for a TVC with a named performer needs the same person in every frame. The tools cannot guarantee that yet.

Accurate camera staging. AI image tools do not understand camera position as a spatial concept. They understand “wide angle” or “close-up” as a visual style learned from training data, not as a 24mm or 85mm lens choice made by a DP. The composition shifts unpredictably between prompts that should produce equivalent coverage.

Lens-specific framing. Related to the above. A 35mm shot and a 50mm shot of the same scene have different falloff, different perspective compression, different distortion. AI tools approximate. They do not solve. For shooting boards the DP will read at pre-pro, this matters.

Timing-aware sequencing. AI tools generate frames in isolation. They do not understand that frame 4 has to set up frame 5’s reveal, or that frame 12 has to land on a music beat at 18 seconds into a 30-second spot. The boarder is solving a temporal problem; the AI is solving twenty independent compositional problems.

On-brief revisions. When the agency comes back and says “the second beat needs to feel more intimate, the product reveal needs to come a second later, and the daughter character should be eight not twelve,” a human boarder makes those changes in an afternoon. An AI tool either re-rolls the whole sequence (losing what was working) or chains a complex ControlNet pipeline (where the time saved versus a redraw collapses).

IP-clear deliverables. This is the one most agencies have only recently started asking about. AI image outputs may include latent reproductions of copyrighted training material. The legal status is unresolved across multiple jurisdictions. For a brand-led TVC, the production company and the agency carry the indemnity risk. A hand-drawn board carries no such risk.

NDA-bound trust. Senior commercial briefs are NDA-bound from the kickoff. The brand cannot have its next campaign sitting in the prompt history of a public AI tool. Enterprise-tier API access mitigates the leak risk but does not eliminate it; the production company carries the assurance risk to the brand.

Hand-readable annotation. A working storyboard is annotated with camera moves, blocking arrows, action notes, and timing cues, scribbled in the artist’s hand. The boards land on a director’s monitor and are marked up further on the day. AI-generated frames are aesthetic objects with no annotation language attached.

Why agencies that tried “AI-first storyboards” in 2025 mostly went back

In the period roughly between mid-2024 and late 2025, a number of UK and US agencies trialled AI-first or AI-only storyboard workflows on real briefs. The pattern most often described in client conversations and in trade-press coverage is the same.

The first weeks felt fast. The first complete brief revealed three problems at once.

The first problem was revisions. The agency creative or the director would come back with notes that a human boarder could action in an afternoon, and the AI workflow would take three days because the entire sequence had to be regenerated to maintain visual consistency. The promised speed evaporated at the first round of amends.

The second problem was the pre-pro meeting. The boards did not survive a real read by a DP and a 1st AD. The lens language was wrong, the blocking did not match the location, and the artist who could have absorbed the notes on the day was not in the room. Productions ended up re-boarding from scratch in the week before the shoot, at premium rush rates.

The third problem was the brand sign-off. Brand clients began asking, in due diligence, whether AI had been used in the commissioned material and whether the agency could indemnify the use. Most agencies could not. The brand pushed back. The production reverted to hand-drawn boards.

By the end of 2025 the conversation across most agency creative and production departments I work with had settled into a steady-state: AI for the reference stage, humans for the boards. Several public commentaries in Campaign, Creative Review, and The Drum across 2025-2026 describe the same arc. I have not seen a published case study in 2026 of a six-figure TVC where AI replaced the storyboard artist end-to-end and the agency presented it as a win.

The hybrid workflow that actually works

The workflow that survived the year is straightforward. It is what most senior commercial visualisers in London now run by default in 2026.

Stage 1 — kickoff and reference. AI tools generate a wide net of style references, mood frames, and look-dev pre-vis. The board artist sometimes runs the prompts; sometimes the agency creative or the director runs them. The output goes into a shared reference deck.

Stage 2 — brief and boarding. The board artist takes the reference deck, the script, and the conversation with the director, and draws the boards by hand. Every frame is hand-composed. The boards are IP-clean by default and can be indemnified to the brand.

Stage 3 — revisions. The board artist absorbs notes from the agency, the director, and the brand. Revisions are made by hand. Same-day amends are standard on agency hours.

Stage 4 — pre-pro. The boards go into the pre-pro room. The artist is reachable through the shoot day in case anything reframes on set.

In this workflow, AI is doing what it is good at — accelerating the reference and exploration phase — and humans are doing what they are good at — composing a sequence, holding character consistency, taking notes, and signing IP-clean delivery. The work the producer pays for is the same. The work the artist does in week one is faster.

Cost comparison: AI vs human boards

The pitch for AI-first boards is that they are cheaper. The reality of total cost of ownership over a typical TVC production cycle tells a different story.

The licence cost of the AI tool is genuinely lower than a board artist’s day rate. The revision cost, the IP risk, and the replacement cost when boards have to be redrawn under deadline pressure usually erase the saving by week three.

The honest version of the comparison: an AI-first workflow can save money on a one-and-done illustration brief where IP risk is low and revisions are minimal. For a brand-led TVC with named performers, brand sign-off, and a six-figure production budget, the hand-drawn board is the lower-risk, lower-total-cost option.

IP and rights: the uncomfortable AI question

This is the part of the conversation most producers want to skip and most brand legal teams want to dig into.

The legal status of AI-generated images in 2026 is unsettled. There is active litigation against Stability AI and Midjourney in the US (the Andersen v. Stability AI class action and related cases) and an ongoing UK High Court case between Getty Images and Stability AI. Outcomes are pending. Several jurisdictions have provisional rules; few have settled rules.

In a brand-led TVC, three risks bear on the AI question.

The first is input risk. The brand cannot have its NDA-bound brief feeding a public AI prompt history. Enterprise API access reduces this but does not zero it.

The second is output risk. AI-generated images may reproduce, in latent form, copyrighted training material. The vendor’s terms of service often disclaim indemnity for this case. The agency and the production company are left carrying it.

The third is assignment risk. Standard commercial production contracts assign all IP in the deliverable to the commissioning party. If a portion of the deliverable was generated by an AI system, the chain of assignment may be incomplete. Brand legal teams have started asking for an explicit “no AI-generated content in the deliverable” warranty on hero campaigns. Agencies and production companies are now signing that warranty for the storyboard stage.

The practical effect for producers in 2026: if the brand asks the agency for an IP-clean indemnity, the agency needs to be able to point at a hand-drawn board. The cheap AI route does not survive the brand legal review on a hero campaign.

What to ask a storyboard artist about their AI use

If you are commissioning boards in 2026, five questions worth putting on the kickoff call.

1. Do you use AI in your workflow? Most working artists do, at the reference stage. A “no, never” answer is unusual in 2026. A “yes, and here is exactly where” answer is what you want.
2. For what specifically? Reference gathering and mood frames are normal. AI-generated deliverable frames are not.
3. Do you disclose it to the agency and the brand? The right answer is yes, in writing, in the project brief.
4. Is the deliverable IP-clean? Can the artist warrant that the boards delivered to the agency contain no AI-generated content? On a hero campaign, you want that warranty.
5. Do you charge differently for AI-assisted time? Most senior artists charge the same day rate regardless of which reference tools they used. If an artist quotes a lower rate “because they use AI,” ask what part of the deliverable that affects.

Seb’s stance on AI in his own work

[SEB POV: confirm — recommend: “I use AI sparingly for reference gathering at the kickoff stage. Sometimes the agency producer sends me a Midjourney mood deck; sometimes I run prompts myself when I want to test a lighting reference quickly. Every frame I deliver is hand-drawn, by me, on a Wacom or on paper. My clients pay for that, and I can sign an IP-clean delivery warranty on every project. When a brand asks the agency ‘is there any AI in this material,’ the agency can put me on the call and the answer is no. That is what the day rate is buying in 2026.”]

What to budget for in 2026

The single most common mistake I see in producer briefs in 2026 is budgeting less for boards because “AI has made it cheaper.” It has not. The work has not got faster on the parts of the job that take time. The kickoff stage is faster. The boarding, the revisions, the pre-pro support, and the indemnity sit at the same cost as they did in 2024.

A working 2026 budget for a 30-second TVC storyboard sits in the same band as it did in 2025: roughly £1,500 to £4,500 for a full B&W board pass with one round of amends, more for colour, more again if the brief includes shooting boards or animatic frames. The rate is not falling. What is changing is which artists can deliver the IP-clean, brand-defensible boards a hero campaign now requires. That is the bar a senior commercial visualiser hits. AI tools do not lower the bar; they raise the question of who can prove they have cleared it.

If you are about to commission boards in 2026, budget the same as you did in 2025. Demand the same quality. Ask the IP question on the kickoff call. The work the brand is paying for is the same work it was paying for two years ago. The conversation around it has just got more interesting.

Sources

1. Wikipedia — Storyboard. https://en.wikipedia.org/wiki/Storyboard (Accessed 2026-05-23). Canonical definition and history of the storyboard as a pre-visualisation tool.
2. Midjourney — Documentation. https://docs.midjourney.com/ (Accessed 2026-05-23). Public capability notes for v6 / v7 image generation including character references.
3. OpenAI — Sora system card. https://openai.com/index/sora-system-card/ (Accessed 2026-05-23). Stated capabilities and limitations of the Sora video model.
4. Runway — Gen-3 Alpha documentation. https://runwayml.com/research/introducing-gen-3-alpha (Accessed 2026-05-23). Published feature set for video and reference-frame generation.
5. Andersen v. Stability AI (US class action). https://en.wikipedia.org/wiki/Andersen_v._Stability_AI (Accessed 2026-05-23). Pending US copyright litigation against Stability AI and Midjourney over training data.
6. Getty Images v. Stability AI (UK High Court). https://www.gettyimages.com/company/newsroom/getty-images-stability-ai-judgment (Accessed 2026-05-23). UK High Court proceedings on generative-AI training data.
7. APA — Advertising Producers Association. https://www.a-p-a.net/ (Accessed 2026-05-23). UK commercial production standards covering deliverable IP and rights clearance.
8. Reuters — WGA / SAG-AFTRA AI safeguards coverage. https://www.reuters.com/business/media-telecom/wga-deal-with-hollywood-studios-includes-ai-safeguards-2023-09-26/ (Accessed 2026-05-23). Industry context on union responses to generative AI in screen production.

About the author

Seb Antoniou is a London-based storyboard artist with 10+ years across Premier League, Bentley, Coca-Cola, Nike and BBC Sport campaigns. He boards for advertising agencies and production companies in the UK and globally, working from script, shot list, or a 20-minute call with the director. About →

Related

• Guide: Storyboard vs shooting board vs animatic
• Case study: Bentley Flying Spur Hybrid
• Service: TV Commercial Storyboards

CTA

Got a project? Tell me about it →