InVideo AI Review 2026: What the Credit Balance Does Not Tell You

Last updated: August 8, 2026

Definition opening and verdict

InVideo AI is a prompt-to-video service. InVideo says it can write, storyboard, generate, and assemble video, with access to 200 or more image, video, audio, and music models. Paid self-serve plans are metered in credits. A separate template editor, invideo Studio, runs under the same brand.

Verdict in one line: The number on InVideo’s plan card is a credit balance, not a video allowance, and on the two top self-serve plans it is a dial. The same plan name spans $170 to $340 a month on Generative, and $900 to at least $9,000 a month on Elite.

Scope: Future Stack Reviews did not buy a plan, generate a video, or measure render speed. This audit examines whether a buyer can price and govern InVideo from its public documents. It does not evaluate output quality, editing precision, reliability, or support.

A credit balance is not a video allowance

On 8 August 2026, InVideo’s individual pricing page showed 75 monthly credits on Plus and 390 on Max, as fixed figures. Generative and Elite were not fixed. Each carried a slider and a dropdown that select the credit allowance at purchase, and the displayed price moved with the selection.

Three things follow, and all three are arithmetic on the figures above rather than interpretation.

Price tracks credits exactly at every observed setting. Doubling Generative’s credits doubled its price. Doubling Elite’s credits doubled its price. Multiplying Elite’s credits by ten multiplied its price by ten. There is no volume discount inside a plan.

Some entitlements scale and some do not. Avatar and voice clone slots scaled with the credit selection, as did Elite’s storage. Concurrency did not: Elite displayed the same twenty times Plus at 4,250 credits and at 42,500.

Upgrading barely changes what a credit costs. Dividing each plan’s annual charge by its annual credit allocation at the base setting gives $0.222 per credit on Plus, $0.214 on Max, $0.208 on Generative, and $0.212 on Elite. That is a spread of under seven percent across a price range from $200 to $108,000 a year, and Elite is fractionally more expensive per credit than Generative. On this measure InVideo’s own “Best value” badge on the Generative card is accurate. What a buyer gains by moving up is storage, iStock, avatar slots and concurrency, not cheaper credits.

Each card also illustrates its balance with Nano Banana image-generation counts, and no card at any setting gave a representative finished-video total. That is the gap this briefing is about. InVideo publishes video rates elsewhere, and those rates belong to different creation paths that cannot be collapsed into one plan-wide exchange rate.

The Plus card’s 300 Nano Banana Pro figure implies 0.25 credits per image, which matches the low end of that model’s published 0.25 to 0.40 range. The card does not say which resolution or quality setting it assumes. The Nano Banana 2 figure implies 0.125 credits per image against a published minimum of 0.12. That 0.005 difference is unexplained in any page reviewed here, and this briefing does not speculate about why.

One last observation belongs in a procurement file rather than a shopping decision. InVideo’s billing help pages list Generative as 800 credits and Elite as 4,250, as single values with no range indicated, and the billing price table has no Elite row at all. A self-serve plan configurable to $108,000 a year does not appear in InVideo’s own published billing price table.

Sources: InVideo Pricing, captured 8 August 2026 · InVideo, AI Image Generation Credits and Usage, opened 7 August 2026

The workflow billing map

A mathematically correct division can still produce a misleading buying claim. Dividing 75 Plus credits by the published Autopilot Basic rate gives 37.5 minutes. That is conditional arithmetic for one stock-only mode. It is not an entitlement to 37.5 minutes of client-ready video, and the same division applied to generative paths would be worse, because those paths charge per model, per second, and per media operation rather than per finished minute.

Sources: InVideo Help Center, Basic, Pro and Ultra, 8 June 2026 · InVideo Help Center, Plans and credits, opened 7 August 2026 · InVideo, AI Video Generation Credits and Usage, opened 7 August 2026 · InVideo Help Center, How credits are charged when using AI Agents, opened 7 August 2026

What the agent can spend before you intervene

InVideo’s agent documentation is more useful for budgeting than any model price table, because it describes the charging event rather than a unit price.

An agent response may reason through the brief, generate images, produce video clips, and design a voice in a single exchange. Each of those operations draws from the same credit balance according to the background models used. The documentation states that the exchange charge can be inspected after the response completes, through the message menu, and that total project spend appears under Project Usage.

Two settings matter before you start. The documentation states that the agent defaults to Always allow, meaning it generates without asking first. A buyer can switch to Ask before generating videos, or to Always ask before generating. The approval prompt shows the prompt, model, duration, and aspect ratio, and the settings can be adjusted before confirming.

The documentation also describes a per-project credit limit. It does not state which self-serve plans include it. Confirm that entitlement in your own account or in writing before relying on it as a spend control.

The practical consequence is narrow and worth stating plainly. On default settings, a self-serve buyer sees the cost of an agent exchange after it has been charged, not before. Changing the permission mode is the single highest-value action a new buyer can take in the first five minutes.

Sources: InVideo Help Center, How credits are charged when using AI Agents, opened 7 August 2026 · InVideo Help Center, Credits calculated before creating, opened 7 August 2026

Limits that are not credits

A paid workflow can also stop at the iStock, storage, seat, or concurrency limit.

iStock. Paid plans include 100 to 5,000 iStock assets depending on tier. This quota is consumed at download, based on how much iStock media the finished video contains. InVideo describes Storyblocks, Shutterstock, Pexels, and Pixabay as unlimited on paid plans, with iStock as the metered library. The agent selects stock media automatically during generation, so keeping iStock out of a project requires an instruction or a limit set before generating.

This produces a wording conflict worth understanding before it surprises you. Two billing pages state that downloading generated content does not consume credits, and one states it as a flat no. A separate help article explains what to do when downloading a video does trigger a payment prompt, and attributes that prompt to the iStock allowance. Both are accurate. The iStock quota is a different meter from credits, and the pages making the absolute statement do not say so.

Storage. Plans include 20 GB to 10 TB. Neither the plan limits page nor the credits pages describe what happens when storage fills. FSR could not locate a documented failure state for storage in InVideo’s billing documentation, and did not test it.

Seats. Individual plans include no additional seats. Collaboration requires the Team plan, which supports two to thirty seats.

Concurrency. Max, Generative, and Elite advertise concurrency at two, ten, and twenty times Plus. Plus itself is described only as limited concurrency. The base value is not published, so none of the three multiples resolves to a number.

Sources: InVideo, Plan Limits, opened 7 August 2026 · InVideo, How Do Credits Work, opened 7 August 2026 · InVideo Help Center, Why am I being asked to pay extra, 8 April 2026 · InVideo Help Center, How is iStock consumed, 18 March 2026 · InVideo Pricing, captured 8 August 2026

What happens if the evaluation fails

The refund terms make the first test consequential, and they are the strongest reason to treat month one as a measurement period rather than production.

InVideo’s Terms of Service, effective 8 July 2026, state that fees are non-refundable for the billing period purchased, with a seven-day exception for an initial purchase or renewal where no credits or premium features have been used. The same section states, “We do not offer pro-rata or partial refunds” for unused portions. Credits are described as non-refundable, non-redeemable, and non-transferable.

The dedicated refund article, dated 12 June 2026, agrees and defines the threshold. A subscription counts as used once any video or asset has been generated or downloaded during the billing period, and it states that “Used subscriptions are not eligible for a refund under any circumstances.”

A third page conflicts. The billing help page that carries InVideo’s plan price table opens by stating, “The refunds are pro-rated and usage based.” Further down, the same page says cancelling does not trigger a refund and that eligibility is reviewed case by case.

The Terms and the dedicated refund article align. One billing page does not. Until InVideo reconciles the wording, budget against the stricter published rule, and assume that generating a single test video can close the standard refund route.

Two channel rules also apply. Purchases made through the Apple App Store or Google Play are billed, cancelled, and refunded through that store rather than through InVideo, and the web billing portal that holds invoices and tax details is not available to app store or UPI subscribers. If a renewal payment fails, InVideo retries over the following days, then downgrades the account to Free, and states that unused paid credits expire at that point.

Sources: InVideo Terms and Conditions, effective 8 July 2026, Section 3(d) · InVideo Help Center, Refunds, 12 June 2026 · InVideo, Subscriptions and Refunds, opened 7 August 2026 · InVideo, Payments, Invoices and Errors, opened 8 August 2026

Money survivorship after cancellation

Outputs and unused balances do not follow the same termination rule.

The Terms grant a perpetual, irrevocable, worldwide, royalty-free, non-exclusive license to use, modify, distribute, and commercialize the outputs you generated, and that license expressly survives cancellation or termination. The same section qualifies it: rights in outputs that incorporate licensed stock content remain subject to that content’s own license terms, and InVideo grants no exclusivity over licensed content or AI-generated variants of it. So the accurate statement is a surviving commercial-use license, not unconditional ownership of everything in the frame.

Unused balances are treated differently, and the naming makes this harder than it needs to be.

The practical rule for a buyer is simple. Before purchasing any credit top-up, ask InVideo in writing which product you are buying and what happens to the unspent balance at your next reset and at cancellation. The published pages do not settle it.

Upgrading mid-cycle carries a separate ambiguity. The pricing page FAQ states that unused credits from the previous plan roll over into the new plan on upgrade. The subscriptions help page states that the new plan’s allowance applies immediately and that credits already used count against it. Those readings diverge for anyone who upgrades after spending most of a cycle, and this briefing does not have an account receipt that would resolve which one the billing system applies.

Sources: InVideo Terms and Conditions, effective 8 July 2026, Section 4(h) · InVideo, Extra Credits, opened 7 August 2026 · InVideo, How Do Credits Work, opened 7 August 2026 · InVideo, Subscriptions and Refunds, opened 7 August 2026 · InVideo Pricing FAQ, captured 8 August 2026

Self-serve and enterprise are different tracks

InVideo’s Privacy Policy, last updated 8 July 2026, separates two contractual tracks in its opening section, and reading that separation first prevents a common misreading of the rest.

Where an organization has signed a Master Services Agreement and submits data as Customer Personal Data, that organization is the controller and InVideo acts as its processor. The Privacy Policy expressly does not apply to data processed solely on behalf of such a customer. Where you sign up directly on the website, InVideo is the controller and the Privacy Policy governs.

That distinction explains why the Enterprise marketing page and the Privacy Policy read differently. The Enterprise page states that inputs and outputs are never used to train InVideo’s models or shared for training with any external models, alongside dedicated isolated environments, purge on request, and commercial-use indemnity. Those are enterprise-track commitments. If you are buying Plus, Max, Generative, or Elite from the website, the Enterprise page is not your contract.

Your contract is the Privacy Policy, and it says two things that a self-serve buyer should read together.

First, InVideo commits that it does not use your User Content, Inputs, Output, or Designs to train or fine-tune its proprietary models. It does use aggregated, de-identified system and usage data to maintain and improve its models and infrastructure.

Second, third-party AI providers are governed by a separate two-tier framework. Standard Tier providers are contractually prohibited from using content for their own purposes, including model training. Experimental Tier is defined as providers that have not committed to refraining from training on Customer Personal Data, or that do not hold complete independent security certifications, and the policy states such providers “may use content for model improvement in accordance with their own terms” and data practices.

The policy directs readers to the subprocessor list for the tier assignments and the jurisdictions involved. On the Trust Center subprocessor page captured on 8 August 2026, twenty-six providers were listed with a category and, for eleven of them, a country. No Standard or Experimental tier label appeared on any entry in that capture. A model-level list does exist elsewhere: the Trust Center FAQ points to InVideo’s models page for experimental models, and that page marks specific models as experimental. The provider-level mapping the Privacy Policy refers to was not located in the pages reviewed here.

For a self-serve buyer processing faces or voices, that mapping is the question to ask in writing before uploading anything sensitive. For an enterprise buyer, the answer belongs in the MSA and DPA rather than on any public page.

Biometric handling itself is documented in more detail than most vendors provide. The Privacy Policy states that facial embeddings and voice characteristics are treated as sensitive or special category data where law requires, processed only with explicit consent captured in a dedicated in-product step, and retained until deletion request or account deletion. Retention periods for other categories are published in a table, including seven years for payment records and twelve months for server logs.

Sources: InVideo Privacy Policy, last updated 8 July 2026, Sections 1, 2(c), 3, 5(a) and 8 · InVideo Enterprise, opened 8 August 2026 · InVideo Trust Center, Subprocessors, captured 8 August 2026 · InVideo Trust Center, FAQ, captured 8 August 2026

Which company you contract with

InVideo contracts through three entities, and which one applies depends on where you are.

The Terms state that mandatory local consumer protection rights continue to apply for users outside the three named jurisdictions.

Three provisions belong in any procurement file.

Total aggregate liability is capped at USD 50, subject to the Terms’ own carve-out for liability that cannot lawfully be limited in your jurisdiction. That cap sits against an Elite annual commitment of $10,800 and against prepaid credit purchases with a stated five dollar minimum and no stated maximum.

International transfers are addressed as a menu rather than a mapping. The Privacy Policy lists the EU-US Data Privacy Framework, Standard Contractual Clauses under Commission Implementing Decision (EU) 2021/914, the UK International Data Transfer Addendum, Swiss-compliant mechanisms, and adequacy decisions. It does not map a specific recipient or data flow to a specific mechanism, and states that a copy for a specific flow is available on request.

The three entities act as joint controllers under GDPR Article 26 for certain data. The essence of that arrangement is available on request rather than published. An EU representative under Article 27 is named.

One change since January runs in the buyer’s favor and should be recorded. The Terms in force on 13 January 2026 applied Indian law and Mumbai arbitration to every user regardless of residence, used a three-arbitrator panel, and placed arbitration costs on the customer. They also stated that InVideo had no obligation to provide support or maintenance, and granted InVideo rights to use AI-based output to improve its artificial intelligence. The July 2026 Terms replace the single forum with three, and add an explicit commitment not to train on user data.

Compliance wording differs by surface and should be read precisely. The Enterprise page states SOC 2 and ISO aligned. The Team and Enterprise pricing tab shows SOC2 and GDPR as a feature checkmark. The Trust Center introduction refers to certifications including SOC 2 and GDPR, while its compliance panel displayed GDPR only in the 8 August capture, and its resources section sits behind an access request. FSR did not obtain or read a SOC 2 report and makes no statement about whether one exists or what it covers. Request the report and its scope statement rather than relying on any of these pages.

Sources: InVideo Terms and Conditions, effective 8 July 2026, Sections 1, 7(c), 7(d) and 10 · InVideo Terms and Conditions, effective 13 January 2026 · InVideo Privacy Policy, last updated 8 July 2026, Sections 1(b), 7 and 14 · InVideo Enterprise, opened 8 August 2026

How to run a one-cycle metered pilot

The public documents cannot tell you what a finished video costs in your workflow. A single billing cycle can. This is the protocol FSR would use, and it is the same protocol that would produce the Tier B evidence this briefing lacks.

FAQ

Sources: each answer restates a claim sourced in the section that discusses it. Primary references: Pricing, captured 8 August 2026 · Terms, 8 July 2026 · Privacy Policy, 8 July 2026 · Agent credit charging

Methodology and source note

This is a Tier C briefing: primary-source document research with no hands-on testing. FSR did not purchase a plan, create an account, generate a video, or observe any InVideo interface in operation.

Every claim above traces to a page FSR opened directly on 7 or 8 August 2026, or to an in-browser capture made by FSR on 8 August 2026 for pages that render through client-side scripts, which include the pricing page, the Trust Center, and the models catalog.

Arithmetic is labeled. The single derived figure in this briefing, 37.5 minutes of Autopilot Basic on Plus, is FSR’s division of a published allowance by a published rate, and it excludes actors, audio, images, iStock, revisions, and failed attempts. No derived figure is a capacity commitment by InVideo.

Absence is scoped. Where this briefing says something was not located, it means it was not found in the pages named in the source lines, on the dates given.

Three exclusions are deliberate. A help article describing a legacy video-minutes allowance was excluded because retrieval was not reproducible across independent attempts on the same day, and a page whose state cannot be reproduced cannot support a present-tense claim. Statements made by InVideo staff on social platforms were excluded because FSR did not open and archive them. Competitor pricing comparisons were excluded because that research was not completed to the standard required here.

The following remain unverified and are not asserted above: the Free plan’s credit allowance, any monthly billed price for Elite, the maximum selectable setting on either configurable plan, iStock and concurrency values above the base setting, the base concurrency value itself, the treatment of failed generations, storage cap behavior, which self-serve plans include a project credit limit, the tier assignment for each named subprocessor, and the contents of any assurance report.

InVideo states that model and agent prices can change without prior notice. Verify current figures on InVideo’s pricing page before purchasing.

FSR has no commercial relationship with InVideo. No affiliate link appears on this page. FSR has not received compensation, review access, or a complimentary account from InVideo.

Recheck trigger: this briefing should be rechecked when InVideo changes its pricing page, its Terms, its Privacy Policy, or its billing help pages, and in any case within ninety days.

Full source list: Pricing · How Do Credits Work · Plan Limits · Video Credits · Image Credits · Audio Credits · Extra Credits · Subscriptions and Refunds · Payments and Invoices · Terms, 8 July 2026 · Terms, 13 January 2026 · Privacy Policy, 8 July 2026 · Acceptable Use Policy · Enterprise · AI Models catalog · Trust Center Subprocessors · Trust Center FAQ · Basic, Pro and Ultra · Agent credit charging · Plans and credits · Credits before creating · Refunds · iStock consumption · Paying extra · Free plan · Understanding credit changes

Corrections ledger

FSR does not correct errors silently. This page was first published on 23 March 2026 and updated on 27 April 2026. That version has been withdrawn. The following claims appeared in it or in an unpublished August rebuild and are retracted.

Verdict and forced decision

InVideo publishes more billing documentation than most vendors in this category, and the July 2026 contract revision improved the customer’s position on training, jurisdiction, and arbitration costs. Those are real points in its favor and this briefing does not dispute them.

What the documentation does not do is convert a plan balance into a finished-video budget. Autopilot charges per minute, direct model generation charges per second, and an agent exchange charges across several media operations at once. The plan card shows a number. The workflow decides what that number buys.

On Generative and Elite the plan card is a configurator rather than a price. The same Elite name covered $900 and $9,000 a month in captures taken eleven minutes apart, and the credit rate barely moved between them, because price and allowance scale together. Anyone quoting “Elite is $900” from a comparison article is quoting the floor.

If you are self-serve: take month-to-month for one cycle and treat it as measurement, not production. Change the agent permission setting before your first prompt. Do not prepay annually and do not buy a large credit top-up from the plan-card examples alone. The refund rules mean the first generated video can close the standard refund route, so the pilot has to be deliberate.

If you are buying for an agency: run the same three briefs twice and quote from credits per accepted output. First-generation cost will understate your real number, because revision depth is where the balance goes.

If you are in procurement or a regulated environment: the Enterprise marketing page is not your contract. Request the MSA, the DPA, the subprocessor tier mapping, the applicable transfer mechanism, and the assurance report scope, and get the credit expiry and refund rules in writing before signing anything annual.

The question this briefing cannot answer is the one that decides everything: what a representative finished workflow actually costs. Public documents do not reveal it. A metered account test does, and that is the next piece of work.