The short answer
There is no single price for an AI film, in the same way there is no single price for a shoot. What has changed is where the money goes. Generative tools remove much of the cost of physically capturing images: locations, crew, kit, travel and build. They do not remove the cost of a good idea, of direction, of finishing, of sound, or of making sure the work is legally safe to run in every market it appears in.
Below is a calm look at what drives cost, what credible sources report for traditional and AI-led work in the US and the UK, what the models themselves charge per second, and how to build a budget that holds.
What drives the cost of an AI film
- Idea and script. The concept, the script and the approvals around them. This shapes everything downstream.
- Direction and look development. The visual world, the casting of AI characters, the camera language and the style frames that keep every shot consistent.
- Generation compute and iterations. Paying the model provider for every clip generated, including the many takes that are never used.
- Editing, compositing and grade. Assembling the cut, fixing artefacts, combining generated and real elements, matching colour and upscaling to the delivery resolution.
- Sound and music. Sound design, mix, and composed or licensed music, cleared for the media and territories you need.
- Voice. A human voiceover artist or a synthetic voice, each with its own consent and usage terms.
- Revisions. Rounds of feedback, and the regeneration they can trigger.
- Legal and clearance. Claims substantiation, likeness and IP checks, and broadcaster clearance where it applies.
- Delivery and versions. Durations, aspect ratios, subtitles, languages and technical specifications for each platform and market.
What traditional production costs
Production budgets are rarely made public, so published benchmarks come mostly from production and media companies themselves. They are useful as orientation, not as quotes. Media spend sits on top in every case and is a separate budget.
United States (USD)
Simulmedia, a US TV advertising company, reports in its 2026 guide that a professionally produced 30-second commercial typically costs between $10,000 and $50,000 to produce. It places national-standard production at $50,000 to $500,000, and celebrity or premium work at $500,000 to $1 million or more. Lemonlight, a Californian production company working in both live action and AI, reports that a professional 30-second brand video can easily reach around $15,000 to $50,000 or more through traditional production.
United Kingdom (GBP)
Lambda Films, a London production company with more than 15 years of experience making TV adverts, reports three bands for production alone in its guide, updated in August 2026: from around £5,000 to £30,000 for simple adverts with minimal locations and cast; £30,000 to £100,000 for professionally shot live action with multiple locations, a small cast and light VFX; and £100,000 to £500,000 or more for larger casts, intricate sets, high-end VFX, bespoke music or celebrity talent.
Across both markets the same drivers recur: locations and travel, talent, crew and equipment, and post-production.
What AI-led work is reported to cost
There is not yet a reliable, independent industry benchmark for AI commercial production in any market. Many figures online come from tool vendors or affiliate sites and cannot be traced to data. Lemonlight's own guide is one of the more transparent published views from a production company: it reports DIY AI tools at roughly $20 to $300 a month, and freelance AI video at around $500 for a simple short-form asset up to $5,000 for a more polished deliverable, with significant variation in quality.
Two widely reported cases show both ends of the spectrum. In June 2025, the prediction market Kalshi aired a 30-second, fully AI-generated spot during a YouTube TV stream of the NBA Finals. NPR reported that the company put the AI prompting cost at under $2,000 and that the creator, advertising veteran P.J. Accetturo, made it in about two days. His own fee was not disclosed. He also offered a caveat worth remembering: just because it was cheap does not mean anyone can do it.
At the other end, TheWrap reported in November 2025 that Coca-Cola's AI-led holiday film involved a team of five AI specialists working through more than 70,000 clips over 30 days, followed by colour work and clean-up. That is not a cheap production in any meaningful sense. It is a different allocation of effort.
What the models charge per second
Raw generation is the one cost that is published openly. These are list prices in US dollars from official pages, checked in October 2026. Local currency billing, volume agreements and resellers can change the effective rate.
- Google Veo 3.1 via the Gemini API (paid tier, with audio): Standard at $0.40 per second for 720p or 1080p and $0.60 for 4K; Fast at $0.10 to $0.12 per second for 720p or 1080p and $0.30 for 4K; Lite at $0.05 to $0.08 per second. Google charges only when a video is successfully generated.
- Runway API: credits cost $0.01 each, and Gen-4.5 uses 12 credits per second, or $0.12 per second. Gen-4 Turbo uses 5 credits per second. Subscription plans run from a free tier to a Max plan at $95 a month billed monthly.
- OpenAI Sora: OpenAI's documentation states that the Sora 2 models and Videos API were shut down on 24 September 2026, so the previously listed rates of $0.10 and $0.30 per second no longer apply. A production built around a single tool carries platform risk.
Why prompt-and-post rarely reaches broadcast grade
Per-second prices look trivial, which is why the cheapest route is so tempting: write a prompt, generate, cut, publish. The catch is the iteration ratio. Describing the Kalshi spot to NPR, Accetturo said it took about 300 to 400 generations to get 15 usable clips, roughly 20 to 27 attempts for every shot that made the edit, on a deliberately chaotic, comedic brief.
As an illustration only: at Veo 3.1 Standard rates, an eight-second take at 1080p costs $3.20. Twenty-five takes per usable shot is $80 a shot, and fifteen shots is around $1,200 in compute. The compute bill is modest. The real cost is the hours of judgement spent choosing takes, holding a character, wardrobe and product consistent across shots, and repairing what the model got nearly right. A polished brand film with tight continuity and a legible product demands far more selection and finishing than a playful social spot.
Generation is also only raw material. The Coca-Cola team described post-production as the new pre-production. Editing, compositing, grade, upscaling, sound design and mix are where generated footage becomes a film.
Where AI genuinely saves money
- No location fees, permits, travel, accommodation, set build or weather days.
- Smaller crews and no kit hire for scenes that never need to be physically shot.
- Faster revisions to a scene, a background or a season, without a reshoot.
- Efficient versioning and localisation: new languages, markets, cut-downs and aspect ratios from one master world.
- Scale and spectacle, such as impossible locations or crowds, that would otherwise sit in the top band of a live-action budget.
Where it does not
- Direction and taste. Someone still has to decide what good looks like and reject hundreds of near misses.
- Post-production. Compositing, grade, clean-up and finishing often take more time, not less.
- Sound. Music licensing, sound design and the mix cost much the same as for a filmed spot.
- Rights and consent. Likenesses, voices and reference material need documented permission, whatever the tool.
- Legal review in each market, outlined below.
Legal review: brief notes by market
- United Kingdom. CAP has said its rules apply regardless of how content is generated, and that a disclosure will not rescue a misleading ad. TV ads go through Clearcast's clearance process before they can air.
- United States. The FTC has stated that there is no AI exemption from the laws on the books, so truth-in-advertising and substantiation rules apply in full.
- European Union. Under Article 50 of the AI Act, deployers must disclose deepfakes as artificially generated or manipulated, with a lighter duty for evidently artistic, satirical or fictional work. These transparency obligations have applied since 2 August 2026.
A practical budgeting checklist for brands
- Define the deliverables first: number of films, durations, aspect ratios, languages and markets.
- Agree the finish required. Social-first, digital out-of-home and broadcast carry very different standards.
- Ask how direction and look development are budgeted, separately from generation.
- Ask for an assumed iteration ratio and a compute allowance, not just a per-second price.
- Check that edit, compositing, grade and upscale are included to the delivery spec.
- Confirm music rights by media, territory and term.
- Confirm voice: human or synthetic, with consent and usage terms in writing.
- Fix the number of revision rounds and what counts as a new brief.
- Allow time for substantiation, likeness and IP checks, and clearance in every market where the work will run.
- Ask which tools will be used and what happens if one is withdrawn mid-project.
- Keep media spend as a separate line.
How we approach budgets at Vane AI
We work with brands worldwide from our base in London, and every project starts with scope, not software. A clear brief and a fixed list of deliverables mean both sides know exactly what will be made, in which formats and for which markets. Direction, generation, post, sound and clearance are planned as distinct stages, so nothing important is assumed to be free. We are open about the tools we use and the assumptions behind the plan, and revisions are agreed upfront rather than discovered at the end.
This article is a general overview as of October 2026 and is not legal advice. Model prices, plans and availability change frequently, and published production ranges are indicative only. Check current pricing on each provider's official page and take local advice for each market before budgeting.