AI video generator pricing follows three models: a monthly subscription with a credit allowance, pay as you go credits, and API billing by the generated second or by the request. What you pay depends on clip length, resolution, model tier and how many takes you discard, so compare tools by cost per finished minute.
How much does an AI video generator cost?
It costs whatever your finished footage consumes, and the plan name tells you little about that. Two tools with the same monthly fee can deliver very different amounts of usable video, because each vendor decides what a credit buys. Rates change often, so read them on the vendor's pricing page on the day you decide.
Three things set your real bill:
The billing unit. Credits, video minutes, generated seconds or a flat charge for each request.
The settings you pick. Longer clips, higher resolution and premium models draw more from the same allowance.
Your keep rate. Every take you throw away was paid for. A team that keeps one take in four pays four times the listed rate for its finished footage.
The third one is the one you control. A clear brief and a tested prompt cut wasted takes more than any plan change, which is why it pays to work from AI video prompt examples and a written creative hypothesis before you spend on volume.
How AI video pricing works: credits, quotas and seats
A credit is a vendor's own currency: you buy or receive credits, and each generation deducts some. A quota is the ceiling on what you can use in a period. A seat is one person's access. Most plans combine all three, and the fine print on each decides your cost.
Credits
Credits are not comparable between vendors. Runway states that each model uses a set number of credits per generation depending on the model, duration and resolution (Runway pricing). HeyGen says the credits a video uses vary with the AI model, the video duration and the complexity of the generation (HeyGen pricing). So more credits does not mean more video. Convert credits into seconds of output on the model and resolution you will use.
Quotas and resets
Subscription allowances reset each billing cycle, and rollover rules differ by vendor and even by plan. Some drop unused credits at reset, some carry them for one cycle, and credits bought as a top-up often follow a separate rule from credits included in the plan. Check two more limits while you are reading: the longest single video a plan allows, and how many generations can run at once.
Seats
Seats multiply the bill for teams. Some tools give each seat its own allowance, others share one pool, and some charge for every extra editor. Some plans add guest access for reviewers, so check before paying for a full seat.
AI video generator pricing compared
The fair way to compare tools is by billing rule first and by rate second. The table shows how five vendors describe their own billing, checked on 2 October 2026. Each link goes to the vendor's current rates.
Tool | What you buy | What changes the cost | Unused allowance |
|---|---|---|---|
Subscription with monthly credits, plus credit top-ups | Model, duration, resolution | Standard and Pro reset each cycle, Max carries up to one month, purchased credits never expire | |
Subscription with monthly credits, paid extra seats on the Business plan | Model, duration, complexity | Rolls over one month on monthly billing, builds up until renewal on annual billing, lost if you cancel | |
Subscription with monthly credits in one shared pool, plus a small allowance of video minutes | Avatars, B-roll, motion graphics and dubbing each count differently | Not stated on the pricing page, so ask | |
Subscription with credits for each seat | Video creation and the generative model used | Does not roll over | |
Pay as you go, billed by the second of video | Model variant and resolution | No allowance to lose |
Protoface belongs in the pay as you go row. API usage is billed in credits, each model is metered either by the duration of its output or as a flat charge for each request, and the rate for every model is listed on the Protoface pricing page. Free monthly credits do not roll over. Purchased credits roll over without expiry.
What a one-minute video really costs
A finished minute costs the rate for every second you generated to get it, including the takes you discarded. Generators produce short clips, so a one-minute video is a sequence of clips, and each usually takes more than one attempt.
Work it out in four steps, using your own figures:
Count the clips. Divide the finished length by the clip length your model produces. Round up. With eight-second clips, one minute needs eight clips.
Multiply by your takes. If you generate three takes to keep one, eight clips become 24 generations.
Convert to generated seconds. 24 generations of eight seconds is 192 generated seconds for 60 finished ones.
Apply the rate. Multiply those seconds by your tool's rate for the model and resolution you chose, or by its credits for each generation.
Three takes is an example, not a benchmark. Measure your own ratio: over one real project, divide the clips you generated by the clips that reached the final cut. A two-minute video doubles every line, and it often needs more takes for each clip, because characters and settings have to stay consistent across more shots.
Compare on finished minutes. Ask every vendor the same question: at the model and resolution you need, how many seconds of output does the plan buy? Divide by your takes per usable clip. That figure, not the credit count, is what you are buying.
Hidden costs: retries, upscaling and watermark removal
Most of the gap between the sticker price and the bill comes from work you did not plan. Check each charge first.
Retries you chose. A clip you dislike is still a completed generation, and it is charged.
Failed generations. Policies differ, so read them. For Veo, Google states that you are only charged if your video is successfully generated (Gemini API pricing). On Protoface, a generation reserves credits up front, a completed run is charged, a failed run releases the reservation, and a request that is rejected outright reserves nothing.
Canceled jobs. On Protoface, canceling before the run starts releases the reservation, and canceling after it starts charges the full reserved amount. Ask other vendors the same question.
Resolution and model tier. Drafting at a low resolution and rendering only approved shots at full quality is the simplest saving available.
Upscaling, audio and lip sync. These are often separate generations with their own charge.
Watermarks. Free and entry plans commonly stamp the output. Synthesia's free Basic plan states that videos carry a Synthesia watermark, for example. Confirm your plan exports clean video before you plan an ad around it.
Is there an unlimited AI video generator plan?
Some vendors sell plans they call unlimited, and the word always comes with conditions. Read which models the plan covers, whether jobs run in a slower queue, what resolution and clip length apply, and whether a fair use policy caps heavy accounts.
How to set a monthly AI video budget
Budget from your content plan, not from the plan tiers. Count the finished seconds you need each month, multiply by your takes per usable clip, add a margin for revisions, and only then look for the plan or credit amount that covers the total.
A worked example for a small marketing team, in generated seconds so it applies to any tool:
Line | How to work it out | Example |
|---|---|---|
Finished footage | Videos a month times length | 6 videos of 30 seconds: 180 seconds |
Production generations | Finished seconds times takes per usable clip | 180 times 3: 540 seconds |
Exploration | A fixed allowance for testing new looks | Set at 120 seconds |
Revision margin | A share of production for stakeholder changes | One fifth of 540: 108 seconds |
Monthly total | Sum of the three | 768 generated seconds |
Swap in your own figures, then price the total at each vendor's current rate.
Separate exploration from production
Exploration spend buys learning: which look fits the brand and which prompt gives usable shots. Production spend buys the approved asset list. Give exploration a small fixed allowance that nobody has to justify, and release production budget only when a direction passes a short check: it matches the brief, stakeholders approved sample clips, and the look can be reproduced from a saved prompt. The same gate applies when you decide when a concept trailer is worth making, and when you split scenes in a hybrid AI and live-action shoot, where a generated shot competes with the cost of filming it.
Set quotas so nobody is stopped mid-project
A shared pool with visible limits works better than a flat cap for each person. Meter the thing that drives cost, which is generated seconds or credits, not the number of clips. Show the remaining balance where people create, warn the budget owner well before the pool runs out, and agree in advance who can approve more and for how long. Protoface reports account spend by day, by operation, by quality tier and by surface, so you can see whether credits went through the API or the studio.
What happens when you run out of credits mid-month?
Generation stops until the allowance resets, you buy more credits or you upgrade. Find out which your tool offers before a deadline. On pay as you go there is no reset: you add credits and continue.
Subscription, pay as you go or API: which pricing model fits
Pick by how steady your output is. Steady monthly volume suits a subscription, uneven volume suits pay as you go, and video generated inside your own product suits an API.
Pricing model | Fits | Watch for |
|---|---|---|
Subscription with credits | One person or a small team with the same output every month | Allowance lost at reset, seat charges, watermark on low tiers |
Pay as you go credits | Campaign work that spikes, agencies billing clients by project | No ceiling unless you set one, so agree a monthly cap |
API billed on usage | Products that generate video for their own users, automated pipelines | Needs a developer, plus rate and concurrency limits |
Custom or enterprise | Large teams that need invoicing, security review or reserved capacity | Annual commitment and a minimum spend |
Is API pricing cheaper than a subscription?
It can be, when your volume is uneven. An API charges only for what you generate, so a quiet month costs little. A subscription wins when you use nearly all of its allowance every month. Compare both at your real monthly total, and count the developer time.
Avatar and presenter video is metered differently
Presenter tools often express their credits as minutes of finished video, and live avatars are billed by session time. For a live avatar, the waste is a session left open with nobody talking. On Protoface, session billing rounds up to the minute, an idle session ends after a set time without inbound audio, and your plan sets how long a session can last and how many can run at once. Start the avatar when the visitor begins the conversation and end it when they leave, and you pay for conversations, not for open tabs.
Common questions
How much does an AI video subscription cost?
It depends on the tier and on what the tier includes: credits or minutes, seats, resolution and watermark removal. Compare plans by how many seconds of output they buy at the model and resolution you need, and read the current rates on each vendor's pricing page.
How much does a 1 minute AI video cost?
Count the clips a minute needs at your model's clip length, multiply by the takes you generate for each usable clip, and apply your tool's rate to the total generated seconds. With eight clips of eight seconds and three takes each, one finished minute is 192 generated seconds.
How much does a 2 minute video cost?
About twice a one-minute video, plus extra takes. Longer videos need more shots that match each other, so the share of clips you discard tends to rise. Measure your own takes per usable clip on a real project and budget from that.
Do you have to pay for an AI video generator?
Not to try one. Many tools have a free plan or trial credits, usually with a watermark, short clips and a small allowance. HeyGen's free plan, for example, limits the number of videos each month (HeyGen pricing). Publishing clean video at volume needs a paid plan or paid credits.
Am I charged for failed or rejected generations?
It depends on the vendor, so read the policy. For Veo, Google states that you are only charged if your video is successfully generated (Gemini API pricing). On Protoface a failed run releases its reserved credits and a rejected request reserves nothing. A clip you simply dislike is a completed generation, so expect to pay for it.
Do unused AI video credits roll over?
Sometimes. Runway resets monthly credits on its Standard and Pro plans, and credits you purchase never expire (Runway pricing). On Protoface, free monthly credits do not roll over and purchased credits roll over without expiry. Check included credits and top-ups separately.
Price your next video before you generate it
Protoface lists the rate for every video model on one page. Take your monthly total of generated seconds and check it against the current rates.
