In 2026 a 30-second commercial can be produced without a camera, a set or a film crew. The cost of generating video has fallen to cents per second, and specialist AI studios now publish their price lists. Yet some of the most expensive AI ads of the past two years were pulled after public backlash, and the cheapest ones rarely sell anything. The difference is not the model. It is direction, selection, editing, sound and the discipline to meet a quality bar.
This guide explains what an AI commercial costs, which tools professionals actually use, where the budget goes, how to brief and contract the work, and when AI video helps or hurts a brand.
What changed in 2026
Three shifts define the market this year.
Generation became raw material. Video APIs cost from about $0.05 to $0.40 per second of output (CloudZero on Veo pricing, TeamDay model comparison). Even at the top of that range, the raw footage for a 30-second spot costs less than a lunch meeting. The value has moved to the people who decide what to generate, which takes to keep and how to cut them.
Vendors proved unstable. OpenAI announced the end of Sora in March 2026, switched off the app on 26 April and the API on 24 September, unwinding a $1 billion deal with Disney along the way (Wikipedia). Kling 4.0 shipped in October 2026 and Seedance 2.5 in July. A studio tied to one model is a studio with a single point of failure.
Aggregators took the self-serve end. Higgsfield passed $1 billion in annualized revenue at a $5.4 billion valuation, and 85% of its usage comes from social media marketers (Sacra). Runway now bundles Kling, Seedance, Wan and Nano Banana inside its own subscription (Runway pricing). Simple “make me a clip” work is becoming a software subscription; directed commercials remain a service.
The model landscape
| Model | Strengths | Limits and risks | Price (reference) |
|---|---|---|---|
| Google Veo 3.1 | Generates dialogue and sound with the video; strong photorealism; Flow for longer projects with persistent characters | Clips up to 8 seconds; SynthID watermark on all outputs | $0.40/s standard ($0.60 in 4K), Fast $0.10–0.30, Lite $0.05–0.08 |
| Kling 3.0 / 4.0 (Kuaishou) | Best price-to-quality; 4.0 adds native 30-second clips, 4K with 10-bit HDR, stereo sound, keyframes | Content filtered under PRC rules | about $0.084/s silent, $0.126/s with sound (720p), $0.42/s in 4K |
| Seedance 2.0 / 2.5 (ByteDance) | Strong for products and ads, precise prompt following; 2.5 accepts up to 30 images, 10 videos and 10 audio clips as references | Cease-and-desist letters from Disney and Netflix; higher IP risk with recognizable characters | $0.07–0.37/s depending on resolution |
| Runway Gen-4.5 / Aleph 2.0 | Quality plus editing tools; studio deals with Lionsgate and AMC | API clips 2–10 seconds, 720p, no sound | about $0.12/s; plans $15, $35 and $95 per month |
| Moonvalley Marey | Trained on licensed data, aimed at broadcast and film | Less versatile than the leaders | Enterprise |
| Adobe Firefly | Positioned as commercially safe; brand-specific custom models | Quality below the top models | Creative Cloud subscription |
Sources: CloudZero, Kling AI, Seedance, Runway, Lushbinary price comparison, Moonvalley, Adobe Firefly.
In practice, a commercial is assembled from many short shots, and each shot goes to the model that handles it best: a product close-up to Seedance, a dialogue scene to Veo, a wide atmospheric shot to Kling. Key frames are often created first as still images (Nano Banana, FLUX or Midjourney) and then animated, which gives far more control over composition than prompting video directly.
What an AI commercial costs
| Option | Typical price | Notes |
|---|---|---|
| Freelancer | $500–5,000 per spot | Quality varies widely; licensing often undefined |
| Specialist AI studio | $1,000–10,000 per spot, usually $3,000–7,000 for 30 seconds | Usually 3–5 variants included |
| Campaign with 6–12 cutdowns | $10,000–20,000 | Hero spot plus social formats |
| “Cinematic” AI production | $20,000–25,000 and up | Complex characters, heavy post-production |
| European AI studio | €1,500–4,500 for 15–60 seconds; €400–800 per extra language | Prices published in euros |
| Traditional production | $15,000–50,000 digital; $50,000–200,000 national TV | Crew, cast, locations, equipment |
Sources: Panda Studios, Green Frog Labs, Mai Dreams Lab, Lemonlight.
Green Frog Labs reports real projects between $3,850 and $16,800 and publishes a sample $3,850 budget: $1,200 for prompt iterations, $580 for generation credits, $700 for editing and clean-up, and $600 for pre-production (source). Credits come to about 15% of the total. That ratio is the single most useful number in this guide: if a quote is mostly “AI credits”, someone is either inflating it or skipping the human work that makes the ad usable.
Where the money really goes: four cases
- Kalshi, NBA Finals. A production budget of about $2,000, 300–400 generations, 15 clips in the final edit, one person working for two to three days — plus a five-figure creative fee on top (Business Insider). The model: a solo director paid for taste and speed, not for compute.
- Coca-Cola, 2025 holiday campaign. Five specialists, more than 70,000 generated clips and about 30 days of work, followed by a wave of criticism (Today).
- McDonald’s Netherlands. According to the studio, 10 specialists worked for seven weeks on “thousands of takes” for a 45-second Christmas spot. It was published on 6 December 2025 and withdrawn days later after backlash (Today).
- Svedka, Super Bowl LX. Described as the first primarily AI-made Super Bowl spot; recreating its Fembot character alone took about four months (TechCrunch).
The pattern is consistent: the bill is for human selection, iteration and craft. AI removes the crew and the set, not the creative process.
The quality bar a professional AI commercial must meet
Buyers have become precise about what they expect. A solid brief and contract should include:
- No visible artifacts on hands, faces and on-screen text.
- Exact product reproduction: logo, packaging, color and proportions match the real product.
- Character consistency across every shot, usually via a custom-trained character or product model.
- A 4K ProRes master and sound mixed to −23 LUFS.
- Cutdowns in 9:16, 1:1 and 16:9.
- Two rounds of revisions, with payment typically split 40/60.
- A commercial license written into the contract, including how long an AI character may be used (for example, 24 months).
These terms come from studios that publish their delivery standards (Mai Dreams Lab). If a provider cannot commit to them in writing, you are buying a demo, not a commercial.
The hybrid production workflow
The 2026 standard is a hybrid pipeline in which AI handles footage and humans handle judgment:
- Brief: goal, audience, channel, offer, mandatory brand elements.
- Script and storyboard: drafted with an LLM, finalized by a human writer and director.
- Look development: style frames and a custom model (LoRA) for the character or product.
- Key frames: still images generated in Nano Banana, FLUX or Midjourney.
- Animation: image-to-video in Kling, Seedance or Veo, shot by shot.
- Editing in Premiere or DaVinci Resolve.
- Voice and music: voice via ElevenLabs, music from licensed sources only.
- Finishing: upscaling (Topaz), color grading, sound mix.
- Compliance: C2PA metadata, visible labels where required, likeness and IP checks.
- Delivery: master, cutdowns, subtitles, license documents.
Music deserves a special warning. Suno is being sued by UMG and Sony and lost a case to GEMA in a Munich court in July 2026 (Wikipedia). For commercial sync, use licensed stock or a service built with the labels, such as Eleven Music (Wikipedia).
How to brief an AI studio
A good brief saves more money than any model choice, because iterations are the largest line in the budget. Before the first call, prepare:
- Objective and metric. Awareness, sign-ups or sales, and the number that will define success (click-through, cost per acquisition, sales lift).
- Channels and formats. Where the spot runs, how long each version is, and which aspect ratios you need.
- Product assets. Reference photos from every angle, logo files, packaging artwork, brand colors and, if available, a 3D model of the product.
- Mandatory and forbidden elements. Claims legal has approved, disclaimers, competitor references to avoid.
- People. Whether you want AI characters, a digital twin of a real spokesperson (with written consent) or real actors in a hybrid setup.
- Music and voice. Licensed sources only, and the languages you need now or later.
- Labeling requirements for each market and platform where the ad will run.
- Approval path. Who signs off on the storyboard, the rough cut and the final, and within how many days.
Realistic timelines follow from the cases above: a single performance spot can be done in days (the Kalshi ad took one person two to three days), while a hero campaign with a custom character takes weeks — about 30 days for Coca-Cola, seven weeks for McDonald’s Netherlands. Budget the time for review rounds, not just for generation.
When AI commercials work — and when they backfire
The evidence on audience reaction is sobering. An Ipsos and Syracuse University study of 3,000 viewers (May 2026) found that only 13% could confidently identify AI-made ads, yet human-made ads were 14% stronger on short-term sales. Advertising executives believed 82% of young consumers felt positive about AI ads; the real figure was 45% (McGraw Hill summary). Around a quarter of Super Bowl LX ads were about AI, and a Harris poll before the game showed consumers were broadly negative about AI advertising (The Drum). Some brands have started using “No AI” as a selling point (DesignRush).
AI video works best where it is judged on performance rather than sentiment:
- Performance creatives: many variants of hooks and scenes for testing, where a weak variant simply gets switched off. See our guide to AI creative testing.
- Product-centric spots: e-commerce, apps and gadgets, where the product, not an emotional human story, is the hero.
- Pre-visualization and pitches: animated storyboards before a live-action shoot; AMC uses Runway for previs and marketing (Wikipedia).
- Localized versions: one master adapted for several markets — covered in our guide to AI dubbing and video localization.
It is riskiest in emotional brand storytelling, holiday campaigns and categories where trust is everything. If the story depends on human warmth, consider a hybrid: real actors and AI-generated environments, or AI for previs and real footage for the final.
Labeling and rights in brief
Platforms now detect and label AI content. Meta automatically shows an “AI info” label on ads detected as AI-made since 1 June 2026 (Meta). TikTok requires an AIGC label or disclaimer in ads (Common Thread Collective). YouTube requires disclosure of realistic AI content and may remove content or exclude channels from its Partner Program for systematic non-disclosure (YouTube Help).
In the EU, Article 50 of the AI Act has required deepfakes to be labeled since 2 August 2026, with fines of up to €15 million or 3% of global turnover, and the Commission’s guidelines say most advertising does not qualify for the lighter regime for artistic content (Reed Smith). In the US, purely AI-generated footage is not protected by copyright, so document the human creative contribution. Our guide to copyright, likeness and labeling covers the details.
Typical mistakes
- Buying “one-click ads”. Without a script, storyboard and director, you get generic footage that viewers recognize as AI slop.
- Vague briefs. A small change can require regenerating a whole scene; unclear briefs burn the budget on iterations.
- One model for everything. Each model has strengths, and any of them can change terms or disappear.
- Recognizable people, characters or brands in references. This is where the IP risk with some models is highest.
- Unlicensed AI music in a paid campaign.
- No license terms in the contract, especially for how long an AI character or digital twin may be used.
- Ignoring labeling until a platform flags the ad or a regulator asks.
- Judging by the showreel instead of by click-through, cost per acquisition and sales.
Conclusion
AI has removed the crew, the set and most of the equipment from commercial production, cutting a 30-second spot from tens of thousands to a few thousand dollars. It has not removed the need for a director, an editor, a sound designer and someone who reads the license terms. The brands that benefit treat AI video as a production method with a defined quality bar, a multi-model pipeline and built-in compliance, and they measure results in sales rather than in views of the making-of.
If you are planning a campaign and want an honest view of whether AI, live action or a hybrid fits your product and audience, book a free 15-minute audit. Our Studio team produces AI commercials with storyboards, licensed music, labeling and a written commercial license by default.