AI video credits are easy to misunderstand because a credit is not a fixed number of seconds, frames, or dollars. It is an internal unit that a platform uses to price a specific generation request. Change the model, duration, resolution, audio mode, or quality setting and the same creative idea may require a different number of credits.
The safest way to budget is therefore not to memorize one “price per video.” Build an estimate from the settings you will actually use, include the tests you expect to reject, and confirm the live credit quote before pressing Generate.
This guide gives you a repeatable method. You can use it while planning a single clip in the imageat AI video generator, a product campaign, or a production workflow that uses several models.
Pricing note: credit values and plans can change. The examples below were checked against live imageat pages on August 25, 2026. Treat the live quote shown in the generation interface as the final source of truth.
What an AI video credit actually represents
A credit is an accounting unit, not a universal measure of compute. One provider might charge a request by seconds, another by output, and another through a credit schedule. Even within one platform, different video models can consume credits at different rates because their upstream costs and available modes differ.
That leads to four practical rules:
- One credit does not equal one second of video.
- A five-second generation is not necessarily half the cost of a ten-second generation.
- Two models can quote different credit totals for the same duration and aspect ratio.
- The number beside the Generate button matters more than an old screenshot, article, or remembered price.
Credits make a multi-model workspace easier to use, but they do not remove the need to inspect each request. Think of the displayed quote as an estimate for one attempt with the current settings.
The five numbers you need before estimating cost
Do not start with your monthly credit balance. Start with the deliverable.
1. Number of final clips
Count the distinct clips you intend to deliver, not just the final edited videos. A 20-second ad made from four generated shots requires four successful clips. If you need three hooks and two product close-ups, that is five deliverables before retakes.
2. Credit quote per attempt
Configure the model, duration, resolution, quality, audio, and other options in the interface. Record the credit number shown for that request. If the platform changes the quote after you switch a setting, use the updated number.
3. Expected attempts per accepted clip
A first attempt is not the same as a usable result. Identity drift, weak motion, malformed hands, unstable labels, or a prompt that simply misses the brief can require another pass.
Use your own acceptance history when available. If you have no history, create a small pilot before budgeting the full campaign. Do not present a guessed acceptance rate as a model guarantee.
4. Number of versions
Aspect-ratio variants, language versions, alternate presenters, hooks, CTAs, and platform-specific edits can multiply generation demand. Some versions can be made in an editor from the same clean output; others need a new generation.
5. Contingency reserve
Keep a separate reserve for genuine failures and late creative changes. A reserve is not permission to generate carelessly. It protects the schedule when one essential shot needs another attempt.
The basic AI video credit formula
Use this formula for one consistent group of shots:
estimated credits = final clips × attempts per accepted clip × credits per attempt
Then add planned variants and a contingency reserve:
total budget = base estimate + variant estimate + reserve
For a campaign with different models or settings, estimate each shot group separately and add the subtotals:
campaign credits = group A + group B + group C + reserve
Do not average unlike requests too early. A short draft, a premium hero shot, and a talking-presenter clip may have different quote structures.
A practical estimation worksheet
Copy this before opening the generator:
Project:
Final deliverable:
Shot group A:
- Number of final clips:
- Model and mode:
- Duration:
- Resolution:
- Audio setting:
- Live credits per attempt:
- Planned attempts per accepted clip:
- Subtotal:
Shot group B:
- Number of final clips:
- Model and mode:
- Duration:
- Resolution:
- Audio setting:
- Live credits per attempt:
- Planned attempts per accepted clip:
- Subtotal:
Required variants:
Contingency reserve:
Total estimated credits:
Current credit balance:
Gap before production:
Save the quote date as well. If production begins later, reopen the project and confirm the settings and displayed cost.
Worked example using a live imageat tool
The imageat AI UGC generator currently displays a request-level quote after you choose the workflow settings. On August 25, 2026, its visible options included Standard at 36 credits and Pro at 54 credits, with five- and ten-second durations. The page also notes that Pro handles non-English scripts noticeably better and that scripts longer than a clip can be cut short.
Suppose you want one accepted product-ad clip in Standard mode and you plan two attempts:
1 final clip × 2 attempts × 36 credits = 72 credits
Suppose the brief genuinely needs Pro mode and you plan two attempts:
1 final clip × 2 attempts × 54 credits = 108 credits
Those are planning examples, not promises about output quality or acceptance. The correct mode depends on the script, language, presenter, product image, and live interface. A 100-credit balance would cover two 36-credit attempts with 28 credits remaining, or one 54-credit attempt with 46 remaining. It would not cover two 54-credit attempts.
This illustrates why “How many videos can I make with 100 credits?” has no single answer. You must define the request and the number of attempts.
Converting credits into an estimated dollar cost
If your plan has a clear subscription price and included credit allowance, calculate an effective plan cost per included credit:
effective cost per included credit = amount paid for the plan ÷ included credits
Then:
estimated generation cost = estimated credits × effective cost per included credit
Keep this number in context. It is a planning allocation, not necessarily a cash charge attached to each click. Subscription features, discounts, taxes, rollover rules, top-ups, and included tools can affect the real economics.
The live imageat pricing page currently states that video estimates use an eight-second reference clip and that actual usage varies by duration, resolution, and audio. It also shows approximate plan-level output estimates. Use those figures for rough plan selection, but use the configured request quote for production budgeting.
When comparing monthly and annual plans, use the amount you will actually pay, not a crossed-out list price. Recalculate whenever the plan, billing period, or credit allowance changes.
Why the estimate changes
Model selection
Video models are not interchangeable compute jobs. They differ in generation architecture, available controls, resolution modes, duration options, audio support, and upstream pricing. Choose the model for the shot, then capture its current quote.
The imageat model comparison hub can help you narrow the creative fit, while the generator provides the request-specific cost.
Duration
Longer output often costs more, but do not assume a perfectly linear rate. A model may offer fixed duration tiers rather than per-second billing. Record the quote for the exact available duration.
Also distinguish generated duration from final edited duration. A 15-second social ad might use three five-second generations plus a static end card. It does not necessarily require one 15-second generation.
Resolution and quality mode
Higher resolution or a premium quality mode may require more credits. Reserve premium settings for shots that benefit from them. Drafting every disposable variation at the highest mode can waste budget.
Resolution does not repair poor composition, incoherent action, or a weak source image. Solve those problems before paying for the final-quality pass.
Native audio, speech, and lip sync
Audio-enabled generation may have a different cost from silent video. A dedicated speaking workflow can also differ from a cinematic text-to-video request.
Decide whether the model must generate audio, whether you will add a voiceover later, or whether you need the dedicated imageat lip sync tool. Do not budget them as the same operation without checking each quote.
Input type and reference controls
Text-to-video, image-to-video, multi-reference, first/last-frame, and video-to-video requests can have different controls and pricing. Starting with a source image may improve control, but a poor source can increase retries.
If you animate a still, use the image-to-video workflow and prepare a frame that already contains the intended subject, composition, lighting, and product geometry.
Variations and batch size
Four variations generally mean four billable attempts, even if they share a prompt. Confirm the displayed total before a batch run. Ask whether you need four full renders or whether four cheap storyboard frames would answer the creative question first.
Estimate a shot list, not a vague idea
The largest budgeting mistake is entering production with “make an ad” instead of a shot list.
For a product video, separate the work into units such as:
- Hook: product enters frame.
- Benefit shot: product used in context.
- Detail shot: texture, packaging, or mechanism.
- Presenter line: short spoken claim.
- End card: real product image, approved copy, and CTA added in editing.
Now label each unit:
- Must generate or can edit from existing footage?
- Draft or final quality?
- Silent, native audio, or lip sync?
- One acceptable version or multiple hooks?
- High-risk interaction or controlled motion?
The product-photo-to-video guide explains how to protect product shape and label details. For social-first ads, the AI UGC video workflow covers scripts, presenters, and platform formats.
A three-pass workflow that wastes fewer credits
Pass 1: Validate the concept cheaply
Use the shortest suitable duration and an appropriate draft or standard mode when available. Test one representative shot. You are checking whether the prompt, framing, source image, and model fit the idea.
Do not generate every variation while the core prompt is still unproven.
Pass 2: Lock motion and continuity
Change one variable at a time. If the camera is wrong, revise the camera instruction without rewriting the entire scene. If the product warps, simplify the interaction and strengthen the source frame.
Review the result frame by frame. The troubleshooting guide to why AI videos look fake can help identify whether the failure comes from motion, identity, lighting, physics, hands, backgrounds, or generated text.
Pass 3: Render only approved shots at final settings
Once the composition and motion work, select the required final duration, resolution, quality, and audio options. Confirm the new credit quote. Generate the final pass, then add exact text, logos, prices, subtitles, and legal copy in an editor rather than asking the video model to preserve typography.
This separates creative exploration from expensive final rendering.
How to reduce retries without lowering quality
Improve the source image first
A clean source frame reduces ambiguity. Keep the subject large enough, remove accidental clutter, use the angle you want to animate, and make sure hands and products are visible rather than partly hidden.
Create or refine the frame with the imageat AI image generator, then use the image upscaler if the source is too small. Upscaling cannot recover a missing angle, but it can prevent avoidable softness.
Give each shot one main action
A person running, turning, opening a bottle, drinking, speaking, and smiling in one generation creates many failure points. Split the sequence. Short controlled shots are easier to regenerate and replace.
Name one camera move
Use instructions such as “locked tripod,” “slow dolly forward,” or “gentle handheld follow.” Avoid stacking orbit, crane, zoom, and fast subject motion unless the shot requires them.
Protect the expensive element
State what must remain unchanged: identity, package dimensions, clothing, logo position, room layout, or camera angle. Keep that element stable and move something less critical.
Add readable text after generation
Changing letters and labels can turn an otherwise usable clip into a reject. Generate clean footage and add approved typography in post-production.
Track real acceptance data
After every project, record:
- Model and mode
- Input type
- Duration and resolution
- Credits quoted per attempt
- Number of attempts
- Number of accepted clips
- Main rejection reason
- Credits spent on accepted footage
Your useful metric is not simply credits per generation. It is credits per accepted clip:
credits per accepted clip = total credits spent ÷ accepted clips
A lower-priced mode that takes many retries can be more expensive than a higher-quality mode that consistently fits the brief. Conversely, a premium mode is wasteful when the concept, framing, or script is still unsettled. Your own project history will answer this more reliably than a universal benchmark.
Pre-generation cost checklist
Before you click Generate, confirm:
- The shot has one clear purpose.
- The correct model and input mode are selected.
- Duration matches the shot rather than the whole edit.
- Aspect ratio matches the destination.
- Resolution and quality are appropriate for this pass.
- Audio or lip sync is enabled only when needed.
- The source image supports the planned motion.
- The prompt names one action and one camera move.
- Required variants are counted.
- The live credit quote matches your worksheet.
- Enough reserve remains for an essential retry.
FAQ
How many seconds of AI video does one credit buy?
There is no universal conversion. The relationship depends on the model and request settings. Check the current quote for the selected duration, resolution, quality, and audio mode.
Why did the credit cost change when I changed models?
Models can have different upstream costs, features, duration tiers, and resolution modes. A platform's credit system translates those differences into request-level quotes.
Is the cheapest model always the most economical?
No. The relevant measure is cost per accepted clip. A low quote can still lead to higher total spend if the shot needs many retries. Test with a small, representative brief.
Should I generate drafts at the lowest resolution?
Use an economical draft setting when it can answer the creative question, but make sure the setting still exposes the failure you are testing. Tiny output may hide face, text, or texture problems that matter later.
Do failed or rejected generations still use credits?
Creative rejection normally does not mean the computation did not happen. Review the live platform terms and request status for the specific case. Budget for iterations rather than assuming every unusable result will be refunded.
How much contingency should I add?
There is no responsible universal percentage. Base it on your historical acceptance rate and shot risk. Until you have data, run a small pilot and reserve enough for at least the essential shots that are most likely to need another attempt.
Can I estimate cost from a monthly plan's “videos per month” number?
Use it only as a rough planning reference because it relies on stated assumptions. The imageat pricing page currently uses an eight-second reference clip and explicitly notes that actual usage varies by duration, resolution, and audio. Request-level quotes are more precise for a real shot list.
What is the fastest way to reduce AI video credit waste?
Validate one representative shot before batching. Use a clean source, one action, one camera move, and the shortest suitable test. Lock the creative direction before switching to final-quality settings.
Final takeaway
AI video cost becomes predictable when you stop asking for a universal credit-to-video conversion and start budgeting specific requests. Count the shots, capture the live quote, estimate attempts and variants, add a deliberate reserve, and track credits per accepted clip after production.
Open the imageat AI video generator, configure one representative shot, and put the displayed quote into the worksheet before generating the rest of the campaign.