In 2026, a raw 60-second AI video costs about $1.80 at $0.03 per second or about $24 at $0.40 per second, before retries. The bill that matters is cost per usable ad, not the monthly plan price.

A rate card tells you what a generation costs. It doesn't tell you how much footage you'll reject, whether reference inputs increase billable duration, or how much editing separates a render from an ad you can publish. Price creative like media: define the unit, track waste, and compare the cost of an outcome rather than the price of access.

AI Video Generator Prices per Second in 2026

2026 AI video rates and clip costs

These are approximate raw model or API costs, before retries, editing, voice work, captions, or legal review. Clip totals are arithmetic estimates at the stated rate, not promises that a model supports that runtime in a single request.

Model or tier	Rate per second	10-second clip	60-second clip
Luma Ray3.2 at 540p	$0.03	$0.30	$1.80
Runway Gen-4 Turbo	$0.05	$0.50	$3.00
Veo 3.1 Lite	$0.05	$0.50	$3.00
Veo 3.1 Fast at 720p	$0.10	$1.00	$6.00
Sora 2 at 720p	$0.10	$1.00	$6.00
Veo 3.1 Fast at 1080p	$0.12	$1.20	$7.20
Kling 4.0 preview, no reference video	$0.157–$0.235	$1.57–$2.35	$9.42–$14.10
Sora 2 Pro	$0.30–$0.70	$3.00–$7.00	$18.00–$42.00
Veo 3.1 standard at 720p or 1080p	$0.40	$4.00	$24.00
Veo 3.1 Standard at 4K	$0.60	$6.00	$36.00

For the table's lower-cost entries, Botmemo's AI video pricing summary lists Luma Ray3.2 at approximately $0.03 per second at 540p, Runway Gen-4 Turbo and Veo 3.1 Lite at approximately $0.05, and full Veo 3.1 at approximately $0.40. The Sora entries follow Beri's video-generation API comparison, which lists Sora 2 at $0.10 per second for 720p and Sora 2 Pro at $0.30–$0.70, depending on resolution.

The Kling and resolution-specific Veo entries use the developer pricing and API comparison cited with their billing mechanics below. Every total in the table assumes you keep all generated footage. That's a budgeting floor, not a finished-ad quote.

Compare configurations, not model names

A buyer asking whether Sora is cheaper than Veo hasn't specified enough. You need the tier, resolution, audio setting, input type, and billable duration before the comparison means anything.

Don't treat the cheapest row as a recommendation. A low-resolution draft may be suitable for judging an opening concept but unsuitable for a close-up product demonstration. Conversely, buying premium generation before you've approved the script pays for fidelity before you've established relevance.

Use this calculation for an initial quote:

Raw generation budget = generated seconds per attempt × number of attempts × model rate.

Then keep post-production separate. A useful video editing fees breakdown helps identify the work outside generation, including assembly, revisions, and delivery.

Creative also needs a media budget. HiveHQ's guide to TikTok ad costs for 2026 is a useful companion when you're deciding how much to allocate to producing candidates versus testing them. Generating more assets isn't automatically valuable if you can't fund a meaningful comparison.

How Subscription Credits and Metered Rates Differ

Sort every quote by what the vendor meters before comparing prices. A subscription, a credit wallet, and a per-second API can describe similar software while exposing the buyer to different kinds of waste.

The practical distinction is easier to see as separate meters.

A comparison infographic explaining three pricing models for AI video generation: subscription credits, metered rates, and flat subscriptions.

Access, tokens, and consumption

Flat subscription: Think of a gym membership. You pay for access and an allowance whether you use it or not, so unused capacity raises your effective cost.
Subscription credits: Think of ride tokens. The token price is visible, but you still need to know how many tokens the actual journey consumes.
Metered API: Think of a taxi meter. Usage determines the generation bill, although minimum charges, input processing, and configuration rules still need checking.

The historical subscription example is instructive: Genie-Artificiel's AI video price history records Runway Standard in 2024 at $15 per month for 125 seconds of Gen-2, equivalent to about $0.12 per generated second when fully used. That's an allowance-utilization calculation, not evidence of a current offer.

Credits add another conversion. Divide the plan charge by its included credits, then multiply by the credits required for your chosen output. Don't use the advertised maximum video count unless it matches your duration and settings.

For teams buying image generation alongside video, AccountShare's guide to compare Midjourney subscription plans provides a separate plan comparison. Keep that purchase distinct from video economics rather than treating every creative subscription as interchangeable capacity.

Old screenshots aren't budgets

Historical prices can mislead even when they were accurate. New Market Pitch reports that Google reduced Veo 3's published price in 2025 from $0.75 to $0.40 per second, approximately 47%.

That change makes a dated screenshot a poor procurement document. Record the quote date, configuration, and billing terms alongside the price you're using.

Before subscribing, write down implied dollars per generated second at full utilization. Treat that as the optimistic case, then calculate what happens when credits expire unused or your team switches to a more expensive setting.

What Duration Audio and Avatars Add to the Bill

Prompt count is a weak predictor of the invoice. A request can consume more billable processing than the duration of the clip you download, particularly when reference footage enters the workflow.

Use the following breakdown to distinguish output length from the other inputs that need a billing check.

A table detailing the costs and multipliers for AI-generated video projects based on duration, audio, and avatars.

Reference footage can change the denominator

Kling's developer pricing document lists no-reference preview rates of approximately $0.157–$0.235 per output second, making a 10-second job roughly $1.57–$2.35; qualifying reference-video workflows use approximately $0.094–$0.141 per billable second, but can bill 20 seconds for that output, producing a total near $1.88–$2.82.

The lower displayed rate doesn't necessarily produce the lower invoice. For qualifying reference jobs, you must apply the rate to processed duration rather than multiplying by downloaded runtime.

Ask for the billing rule in plain language: what input duration is counted, what output duration is counted, and how the vendor combines them. Don't accept “discounted reference generation” as a complete answer.

Resolution needs its own business case

The following table calculates raw clip costs from JSON2Video's API comparison, which lists Veo 3.1 Fast at approximately $0.10, $0.12, and $0.30 per second for 720p, 1080p, and 4K respectively, and Standard at approximately $0.40 for 720p/1080p and $0.60 for 4K.

Veo 3.1 cost for one 8-second clip

Veo 3.1 setting	Rate per second	8-second clip
Fast at 720p	$0.10	$0.80
Fast at 1080p	$0.12	$0.96
Fast at 4K	$0.30	$2.40
Standard at 720p or 1080p	$0.40	$3.20
Standard at 4K	$0.60	$4.80

Reserve higher-resolution renders for work that needs cropping, reframing, or reuse. Evaluate composition and the offer with draft settings before paying to improve detail.

For a visual companion, use this video while reviewing the generation workflow.

Audio and avatars need a separate scope check rather than an assumed universal surcharge. Ask whether synchronized speech, custom voices, avatar access, and cloning are included, metered, or sold as separate entitlements.

Generate modular components when possible. If only the opening needs changing, replace the opening rather than regenerating an approved demonstration and call-to-action.

How to Compare Credits Fees and Usage Limits

Choose a reference job before opening pricing pages. Specify the same scripts, aspect ratio, output duration, voice requirements, reference assets, and acceptance criteria for every vendor.

For example, use a batch of vertical product ads with spoken narration and captions. Keep the source material identical so you're comparing delivery economics rather than different creative assignments.

An infographic showing six steps to compare AI video generator pricing using a standard reference job.

Run the same procurement checks

Convert credits to the required output. Request the credit consumption for your exact model, resolution, duration, and audio configuration. If the vendor won't disclose the conversion, you can't produce a defensible budget.
Establish who pays for failures. Ask about technical failures, moderated requests, interrupted renders, and successful renders that don't meet your requirements. These aren't necessarily treated alike, so record the policy for each.
Inspect input-related charges. Have the vendor price reference footage, lip-sync, and translated versions explicitly. Don't assume that adapting an existing clip is billed like generating a fresh one.
Separate permission from generation. Confirm commercial rights, watermark removal, export access, and download expiry. An inexpensive render isn't a usable business asset if the required rights or delivery options are missing.
Test delivery constraints. Check concurrency, queues, monthly caps, and what happens when you exhaust the allowance. A launch team needs to know whether it can buy more capacity or must wait.
Identify account-level charges. Price additional editors, client access, custom avatars, and voice clones against the actual staffing plan. Don't let a single-user demonstration stand in for an agency workflow.

Define usable before judging cheap

Write the acceptance criteria before testing: product appearance, intelligible speech, accurate claims, readable captions, correct framing, and export quality. Otherwise, reviewers can change the standard without anyone noticing to favor whichever tool they prefer.

Use a shared review sheet with a rejection reason for every discarded output. An iterative testing workflow helps turn those reasons into changes to scripts, prompts, and production instructions rather than repeated blind retries.

The cheaper plan is the one that delivers the accepted reference job for less, not the one with the smallest monthly charge.

Keep creative quality and operational reliability in the comparison. A tool that needs extensive supervision may still be worth buying, but that supervision belongs in its cost.

Monthly Run Rates and ROI for Ad Teams

Monthly creative test costs

The raw-generation rows below are illustrative calculations using the cited rate card, not forecasts of acceptance rates. The production rows identify information you must obtain rather than substitute an unsupported benchmark.

Test plan	Volume	Rate used	Estimated cost
Short hooks on a cheap tier	10 clips of 10 seconds	$0.03 per second	$3
Same hooks on a premium tier	10 clips of 10 seconds	$0.40 per second	$40
Hook matrix on Fast 720p	50 clips of 8 seconds	$0.10 per second	$40
Same matrix on a standard tier	50 clips of 8 seconds	$0.40 per second	$160
Finished UGC clips, vendor-cited pricing required	50 finished videos	Confirm current finished-video quote	Quote × accepted videos
Creator or studio retainer	Ongoing production	Written scope and monthly quote	Retainer plus out-of-scope work

Don't put an unsupported finished-video price beside API arithmetic and call it a comparison. A finished-video quote needs a definition of duration, revisions, voice, captions, rights, and acceptance.

The same discipline applies to studio retainers. Compare the contracted deliverables and production responsibilities, not a generic monthly range.

A visual guide explaining monthly run rates and ROI calculations for advertising creative content generation projects.

Build an acceptance-adjusted budget

Track all billed attempts, not just exported files. When a team retries an opening, changes the product angle, or discards a visual artifact, assign that expenditure to the concept that required it.

For a monthly production ledger, record:

Generation spend: Charges for every billed render, including rejected attempts.
Completion work: Editing, captions, voice correction, resizing, and review.
Accepted output: Assets that meet the publishing specification, not merely downloadable files.
Validation spend: Media used to determine whether accepted ads perform.

Calculate cost per usable ad by dividing production expenditure by accepted ads. Separately calculate cost per validated winner using an agreed performance threshold and including the test expenditure required to find it.

Those measures answer different questions. A production team can deliver inexpensive, technically acceptable ads without finding a persuasive offer or a compelling opening.

Don't mistake cheap production for ROI

Lower production expense improves the cost side of the equation. It doesn't establish that customers will buy, install, or subscribe.

Keep the revenue calculation explicit with a consistent ROAS calculation method. Decide whether your internal profitability view includes production expense, then apply that convention consistently across AI, creator, and studio workflows.

A cheaper model may justify broader exploration. A more expensive model may justify itself if it requires less corrective work or produces footage suitable for a demanding placement. Neither conclusion follows from the rate card alone.

Budgeting rule: Buy enough production capacity to support a test you can evaluate, not enough output to fill a dashboard.

Choose the run-rate row closest to your intended workload, then replace theoretical perfect acceptance with your own observed results. Leave rejected concepts in the ledger. Removing them makes the next budget look better without making the next campaign cheaper.

Subscription Choices and Negotiation Levers

Default to monthly billing until you've completed a representative production cycle. Annual prepayment may reduce the access price, but it also commits you before you've tested the workflow under ordinary review and revision pressure.

Treat the purchase as a production contract. Negotiate the terms that determine usable capacity before asking for a discount on the sticker price.

Trade for operational certainty

Ask for credit rollover when your production schedule is uneven. Request higher concurrency when deadlines matter more than total monthly output.

Get the overage schedule in writing and ask whether the vendor can preserve the quoted conversion when models or plan names change. Bonus credits aren't a substitute for a published consumption table.

Control premium access: Let approved producers select expensive settings; keep exploratory work on draft configurations.
Buy reusable identities deliberately: Add custom-avatar or voice-clone access only when the campaign plan calls for recurring use and the necessary permissions are documented.
Set exit conditions: Define cancellation timing, export access, and what happens to unused purchased capacity.

Separate production from adjacent services

Agencies should distinguish internal production expense from client pass-through charges. Confirm that the rights cover each end brand, and document which revisions belong inside the quoted scope.

Use a content creator rates guide to structure the human-production side of that comparison. Creative direction, filming, licensing, and editing aren't interchangeable with generated footage.

Adjacent services also need their own budget. If you're considering retention services from Class A Media, evaluate their deliverables separately rather than comparing a broader service engagement with a generation allowance.

Commit more volume only when accepted-output cost and delivery reliability remain stable across representative work. A discount is valuable after you've verified the unit you're buying.

Trial AdCrafty When Unit Cost Must Stay Clear

Use a raw API when your team wants direct model selection, can implement billing controls, and already has an editing and review pipeline. That route suits buyers who want to manage generation as a production component rather than purchase an assembled workflow.

Consider a finished-ad platform when the main requirement is moving from a product brief to a publishable asset with fewer handoffs. Evaluate the entire workflow, but don't assume that bundled tools eliminate retries or review.

Give the trial a pass-or-fail test

Start with an approved script and product assets. Use the same acceptance criteria you'd apply to an API-generated ad, including accurate product depiction, clear speech, and usable export quality.

Record the credits spent on accepted and rejected outputs. Include the time required to correct captions, replace footage, and assemble the final asset.

Verify the current offer: Confirm subscription price, trial conditions, included credits, and consumption rules before starting.
Inspect model-specific costs: Check whether changing the model, duration, audio, or resolution changes credit usage.
Test delivery: Export the asset and review it in the intended placement, not just the platform preview.

Choose the workflow, not the headline

Keep the platform if it delivers your reference job at an acceptable total cost and with manageable review effort. Stay with direct APIs if their control and integration fit your team better after completion work is included.

Don't extrapolate a successful demonstration into a guaranteed monthly output. Use the trial to establish an observed cost per accepted ad, then test whether that result survives different products, scripts, and reviewers.

AdCrafty offers an AI-powered UGC creative workflow as an alternative to assembling generated footage through separate tools. Visit the platform to confirm its current plans and trial terms, then run your reference job and record credits, rejected outputs, and editing time before committing.