D-ID

By Minel Gunesoglu, founder. On 27 August 2026 I read D-ID's Studio pricing page in a browser with the billing switch thrown both ways, went through the Studio End User License Agreement clause by clause, checked the vendor FAQ against it, and read two customer-review samples of 131 and 39 that label their own collection differently. Nothing here rests on a video: none was generated for this page, and no avatar was made. Published 28 August 2026; re-checked monthly.

D-ID's Studio pricing page advertises a paid plan at $5.90 a month on monthly billing. The Studio End User License Agreement, section 20.2, licenses that same plan for non-commercial use only. Both sentences were read on 27 August 2026, and the distance between them is what this page is about.

The practical consequence is a single number. If the video is going on a client's channel, into an ad, behind a paywall or anywhere else that earns money, the first D-ID plan that permits it is Pro at $29 a month on monthly billing. The $5.90 monthly figure is real. It is just not the price of a video you may sell.

D-ID Pricing by Plan: What Each Tier Costs Monthly and Annually

Everything below is the Studio surface, read from d-id.com/pricing/studio/ on 27 August 2026, in USD. If you came here for the per-minute cost of D-ID's API, this is not the page: that product is sold on a separate sheet, this desk read it in one billing view only, and no API cost figure appears anywhere below.

The billing basis matters more here than on most pricing pages, because the page opens on the annual view. In that opening annual state the cards show $4.70, $16 and $108 per month, each with an annual total underneath. The control that changes it is labelled "Choose your billing cycle" and sets Monthly against "Annual Save up to 45%". Both states were clicked and read. The monthly-billed figures are the ones this card scores, because those are the ones with no twelve-month commitment attached.

The minute allowance is not video-only. Lite's line reads "10 min. / month for Videos, Agents, Video Translate & API" and Pro's reads "15 min. / month for Videos, Agents, Video Translate & API", so one pool covers the studio, the conversational agents, the translation feature and API calls. Automating does not buy a second allowance. It spends the first one.

Three annual totals round in the vendor's favour against the monthly-equivalent figures printed beside them. Multiplying those annual-view per-month figures by twelve gives $56.40, $192 and $1,296, against printed annual totals of $56, $191 and $1,293. The multiplication is mine; the six figures it uses are D-ID's, off the same screen on the same day. The gaps run from cents to a few dollars a year, and they are here only so a buyer who does the arithmetic knows why the two numbers disagree.

Is D-ID Free? The 14-Day Trial, and What It Is Licensed For

D-ID has no ongoing free plan. The only $0 offering is a 14-day trial carrying 3 minutes of video, a full-screen watermark, and a licence that section 3.1 of the Studio agreement restricts to "with limited features for non-commercial use only", read 27 August 2026. When the fortnight ends, so does the access.

That is worth stating plainly, because the trial beats the cheapest paid plan in one specific way covered further down: on the vendor's own FAQ, the trial reaches a resolution tier that Lite does not.

Why D-ID's Lite Plan Shows Two Different Monthly Figures, $4.70 and $5.90

Several figures for the same D-ID plan circulate, and the reason is almost always an unlabelled billing basis rather than a change in what the vendor charges. G2's D-ID pricing page prints a summary table reading "Lite — $4.70 / Per Month", which is the annual-billed equivalent rather than the monthly-billed figure of $5.90. The same page does disclose the basis lower down, where the plan card states "$56 billed annually (20% discount)" and the summary says "Paid plans start at $4.70/month (billed annually)". So the table is a hazard for a skim-reader rather than a false statement, and G2 notes on that page that "Pricing information is supplied by the software provider or retrieved from publicly accessible pricing materials".

Two freshness stamps sit on that one G2 page, and either could be quoted as the date: the header reads "Last updated on Apr 30, 2026" while the pricing block below closes with "Pricing information was last updated on January 09, 2026". Both were read on 27 August 2026.

The rule that makes all of this tractable: never accept a D-ID figure without the word monthly or annually attached to it.

Can D-ID's Lite Plan Be Used Commercially? Section 20.2 of the Studio EULA

No, and the restriction sits in the licence agreement rather than only in a feature list. Section 20.2 of the Studio End User License Agreement, read 27 August 2026, closes with this:

It is hereby clarified that (notwithstanding any other provision of this Agreement and to the extent applicable) Paying Users subscribed to the “ Lite ” Plan will be able to use the Software and/or Animations for non-commercial use only.

The document's own spacing and curly quotation marks around Lite are reproduced above, because that is how the agreement writes it. The footer of that same licence agreement carries the stamp "Last updated: July 2024".

The pricing page says the same thing in three words without ever using the word restriction. On the Studio pricing page the phrase "Commercial use license" first appears in the feature list of the Pro plan, at $29 a month on monthly billing, sitting among things a buyer gains rather than among things Lite lacks. Read the two documents together and the conclusion is one sentence: Lite is a plan for personal projects, and Pro is the entry point for work you intend to publish or invoice.

The same shape appears on D-ID's separate API price sheet, where the cheapest paid plan lists "Personal license" and the tier above it lists "Commercial license", both read 27 August 2026. That sheet is priced independently of Studio and this desk read it in one billing view only, so no API cost figure is published anywhere on this page. Where a Studio tier includes API access, the plan card above says so.

What a Paid D-ID Plan Says About Who Owns the Video

Ownership on a paid D-ID plan is unambiguous, and it is good news. The same section 20.2 states that "you retain all of your ownership rights in your User Submissions, Output and/or Animations". Nothing in the agreement takes a share of the finished animation back.

Two qualifications travel with it. The agreement takes a broad licence over what you upload, in order to "enable inclusion and use thereof as contemplated by this Agreement", and section 14 runs the indemnity one way: "You agree to defend, indemnify and hold harmless D-ID and our affiliates" against claims arising from your use of the software and from what you submit. That direction is the ordinary position across the vendor agreements this desk has read, so it is context rather than a mark against this product in particular. Section 13.2 caps the vendor's own aggregate liability at what you paid in the preceding twelve months, and section 26 sets Delaware law and Delaware courts for US residents, Israeli law and Tel Aviv courts for everyone else.

Does Any Paid D-ID Plan Remove the Watermark? The FAQ and the Licence Agree

No plan removes it, at any price, and the two documents that say so do not contradict each other. The FAQ, read 27 August 2026, publishes the ladder: "Trial and Lite plan get a D-ID logo watermark.", Pro and Advanced move to a generic AI watermark, and "Enterprise users can customize the AI watermark but not remove it."

The licence is the stronger source, and it resolves an apparent conflict. Section 4.2 defines two different marks: "All Animations contain a synthetic mark" on every output at every tier, while animations made for trial users carry additional watermarks on top of it. So section 3.2's promise that animations produced for paying users "will not have Watermarks" withholds the extra trial marks, not the synthetic one. Section 4.2 then does the work the FAQ only describes, forbidding users to "remove, hide or minimize the Synthetic Mark and/or any Watermarks without obtaining D-ID’s prior written approval", and offering Enterprise customers "Replacement of the D-ID logo with an alternate watermark" subject to written approval. Replacement, not removal.

D-ID gives its reason on the same FAQ page: transparency about the synthetic nature of the videos its users generate, with a link to its published ethics statement. That is a stated position, and this page records what a buyer receives rather than why.

The practical read for a paying customer: at $29 a month on monthly billing the video is yours, you may sell it, and it still carries a mark identifying it as generated.

How to Cancel a D-ID Subscription, and What the Refund Terms Say

Two D-ID surfaces answer the cancellation question differently, and the licence agreement carries the qualifier that decides it. The FAQ's billing section says only "You can cancel your monthly subscription any time". Section 8.2 of the same vendor's agreement adds a deadline:

In order to avoid being billed for the next Billing Period’s Subscription Fee, you must cancel your automatic subscription fifteen (15) days before the end of the applicable Billing Period according to your Plan.

Both were read on 27 August 2026. You can indeed cancel at any moment. Whether that cancellation stops the next charge depends on which side of a fifteen-day line you are on, and only one of the two documents mentions the line.

What happens on the wrong side of it is section 8.5: "Payments shall be made without any right of set-off or deduction and are irrevocable and nonrefundable (except as explicitly provided herein)." The carve-out was traced through the whole agreement and resolves to nothing. There is no "we will refund", no money-back provision and no pro-rata clause; the only other appearance of the word refund is an Apple App Store passage in which Apple, not D-ID, issues it. That is a firmer no-refund position than the "except where required by law" wording several agreements in this corpus carry.

A Trustpilot reviewer describes meeting exactly that wall after a renewal on 19 November 2025:

in the fine print, it says that once a membership auto-renews it will not be refunded.

The document and the lived report arrived here separately and agree.

The mechanics are on the FAQ, in the vendor's own words: cancel on the “Account & API” page, reached by clicking the menu at the bottom left, then “Account & API” > Plan and billing > Cancel Plan, and mobile buyers unsubscribe through their store settings instead. The same page answers what happens to what you already bought: “Your remaining credits will remain valid until the end of the current billing period.” So the credits survive the cancellation, and only the next charge is at stake. If you are trying D-ID with one project in mind, set a calendar reminder fifteen days ahead of your billing date rather than planning to cancel when the work is done.

How D-ID Credits Are Spent, Including on Generations That Fail

D-ID meters output in credits as well as minutes, and the FAQ publishes the conversion: "Each credit is worth up to 15 seconds of video", with longer videos adding up, so that "a 40-second video consumes 3 credits". Each plan sells three package sizes: Lite's cards list 40, 52 or 64 credits, Pro's 60, 100 or 240, and Advanced's 400, 600 or 700, all read 27 August 2026.

The mechanic that matters to a buyer is in none of those numbers. Credits are drawn against the attempt rather than the result, and that shows up at both ends of the star scale across three years of reviews. A five-star Trustpilot review dated 2 August 2023 names it as the product's single flaw:

The only flaw is the credit system; if the video turns out poorly, credits are still deducted

A review dated 28 August 2024 puts the same mechanic in terms of what cannot be done afterwards:

Once a video gets generated your credits are gone. There's no going back and editing the video for free.

And a one-star review dated 22 December 2025 reports it around a failure rather than a disappointment, with the reviewer's own spelling kept:

The Video Avatar failed to generate after multiple tries and comsumed the credits

A complaint that a satisfied customer and an angry one both raise, unprompted and two years apart, reads as a property of the credit system rather than one bad week. Budget for retries. The plan a buyer needs here is the one that survives the takes they throw away, not the one whose headline minutes match the finished runtime.

Do Unused D-ID Credits Roll Over? Section 8.2 Answers It

They do not, and both documents say so. The FAQ is flat: "Credits do not accumulate, they are renewed every month and unused credits become void." Section 8.2 of the agreement puts it in contract terms, that "Unused Credits will expire at the end of the applicable Billing Period and would not be rolled over to a subsequent Billing Period."

An unused D-ID month is therefore a spent month. For a creator whose work arrives in bursts, that is a real cost of the subscription model, and it is the one thing a bigger plan cannot fix.

D-ID Video Quality by Plan: Which Presenters and Resolutions Each Tier Reaches

Output resolution on D-ID is tiered twice over, by plan and by presenter class, and this is the hardest fact on the product to find. It is not on the pricing page, not on the features page, not on the technical-specifications page, and not in the API reference at docs.d-id.com. It is on the FAQ, read 27 August 2026, which says that "Standard AI Presenter output resolution is up to 1280×1280 pixels on all plans", then splits the premium presenters by plan: "Lite plan – Premium presenters not supported" against "Trial, Pro, Advanced and Enterprise plans – 1080p".

Read the ordering twice, because it runs the way nobody expects. Lite, the cheapest paid D-ID plan, is the one named tier that cannot reach 1080p. The free trial can. A Lite subscriber is capped at a 1280 by 1280 square, roughly four-fifths of a 1080p frame's pixel count, and a square is the wrong shape for most places a video ends up.

G2's D-ID pricing page carries the same gate in the vendor's own supplied feature list, where the trial card reads "Premium & Standard Avatars" and the Lite card reads "Standard Avatars only". Two surfaces, one restriction.

Everything else about the deliverable is flat across plans. "All videos are generated in MP4 format", which is one finished file that opens in any editor rather than a project you can reopen inside the Studio, and "the video length is limited to 5 min." per generation on both the studio and the API.

Where D-ID's 68.1 Comes From, and Which Row Does Not Answer Its Own Question

The card was computed on 27 August 2026 and its confidence band reads average.

The value row carries the finding at the top of this page. It is scored on the $29 monthly-billed Pro plan rather than the $5.90 monthly-billed Lite one, because Pro is the first plan a commercial buyer can actually use, and that choice follows a rule this desk already applies to other products whose entry tier is licensed for personal use only.

Commercial terms reads four clause positions rather than a reputation: the one-way indemnity, the paid-tier ownership grant, the non-commercial entry tier and the no-refund posture. What it records is where those four clauses stood in the documents on the day they were read. Whether the company later acts the way its own contract permits is a different question, and this row is not built to answer it.

The top row needs a warning and the card carries it: that row does not answer the question its axis is named for. Nothing public grades how good an avatar video looks, and Artificial Analysis was checked by name for D-ID on 27 August 2026 and runs no board for talking-head generation. So rather than publish a hole, the card publishes a different measurement under its own name, Delivery envelope, answering a question a buyer can act on: what does the entry paid tier let you take away? Because one of the four rows carries a named criterion instead of its own axis, the composite is not a plain average of four like-for-like readings and does not line up cell for cell against a card whose every row measures what it is named for.

What the Delivery Envelope Row Reads on a $29 D-ID Plan

Four published positions, all read on the Pro tier at $29 a month on monthly billing, the same tier the value row prices.

Fidelity reaches 1080p, through the premium presenters Pro unlocks and Lite does not. Format is a finished MP4: portable everywhere, re-editable nowhere. Watermark is present, and the licence rather than the support desk is the source for that. Automation is documented and included, since Pro's own minute pool covers API calls and the public reference at docs.d-id.com served its endpoint documentation to an anonymous request on 27 August 2026, so no sales conversation stands between a subscriber and the API.

The watermark is what holds this row down. Read the row as a description of the deliverable rather than as a verdict on how the avatars look, because it is not measuring that.

D-ID Reviews on G2 and Trustpilot: Two Samples Collected Two Different Ways

Two platforms carry D-ID records of a usable size. Both were read in a rendered browser on 27 August 2026, because both refuse scripted requests, and they disagree by more than half a star.

On G2 the figure is 4.6 out of 5 across 131 reviews, and every review transcribed from G2's own listing carries the collection label "Source: Seller invite" beside an incentivised tag, so this is a sample the seller asked for. Its published star split of 104, 17, 8, 1 and 1 sums to the same 131 and averages, unweighted, to 4.70 against the published 4.6. That arithmetic is mine; the counts are G2's.

On Trustpilot the figure is 3.9 out of 5 across 39 reviews, from a profile the company claimed in June 2026. Trustpilot prints its own caveats on that page, "No recent history of asking for reviews" and "This company hasn't invited customers recently, so reviews may not be representative", and every review on the Trustpilot profile is tagged "Unprompted review".

Two more numbers from that same Trustpilot screen belong on the record, because they explain a gap rather than accuse anybody of one. Its published star split reads 31 per cent five-star, 5 per cent four-star, none at three, 3 per cent two-star and 61 per cent one-star, which averages unweighted to 2.42 rather than 3.9. The platform's own linked explanation is that its score weights recent reviews above old ones. Fourteen of the 39 fall in the last twelve months. Of the reviews read in that session, every one dated June or July 2026 is positive and every one dated inside 2025 is negative. Those are the dates and the counts. No reading of why is offered here, because none is measurable from what is on the page.

This card's usability row rests on the G2 sample alone. Two records collected under opposite regimes are not one sample, so the engine refuses to pool them, and the smaller Trustpilot record ships here in prose with its own figure and its own count instead of being averaged into a cell. The row reads 87.9 rather than a straight conversion of 4.6 out of 5, because a 131-review sample is pulled toward the middle of the scale.

Two Reddit threads surface for this query, and both JSON endpoints returned 403 to this desk on 27 August 2026, so nothing from them is quoted here. That is a gap in what could be read, not evidence that the threads say nothing.

What D-ID Customers Report About Support and Failed Generations

Three Trustpilot reviews between January and August 2025, none referring to the other two, describe the same pair of problems: a generation that will not complete, and a support channel that does not resolve it.

The platform makes it nearly impossible to create a usable avatar — it gets stuck at 99%, crashes, or just doesn't process.

That one is dated 19 May 2025. A review dated 4 August 2025 reports the second half:

Have tried multiple times to contact them and the people who answer haven't got a clue what they're talking about

And one dated 21 July 2025 raises account closure rather than generation:

You cant delete account. They will spam you with garbage messages all the time.

The FAQ's own answer on deletion is consistent with the difficulty described: deleting a D-ID account goes through an email to support, while mobile users can do it in their store settings. Set against that, G2's own summary of its 131 reviews names ease of use, avatar realism and pricing as the recurring themes and does not raise support responsiveness. The two platforms are not shown to disagree so much as the invited sample does not surface what the uninvited one does.

One older review is worth keeping for what it says about the shape of the entry plan rather than about its cost. Dated 17 March 2024, it reads: "For $5.99/mo you only get 10 minutes of video a month?" The figure has moved nine cents since. The ten minutes and the watermark on the cheapest paid plan have not.

D-ID Alternatives That Carry Their Own Dated Score Here

The other avatar tools scored on this site have been through the same rows on their own dates and their own evidence: Synthesia, HeyGen and Hedra. Their order on the avatar generators hub is recomputed from the scores on every build, so nobody chooses who sits where and no sentence here needs rewriting when one of them is rescored.

One directional fact belongs beside those links, and it is a sentence rather than a table: D-ID's $5.90 headline undersells its own commercial floor, and that floor, $29 a month on monthly billing, is the same figure the entry paid tier carries on the Synthesia and HeyGen cards published here, each read on its own date and recorded separately. What follows from that is narrow. It makes none of them better or worse than the others. It means a buyer who picked D-ID on the strength of the advertised entry price was picking on a number that does not apply to commercial work.

Read the top row on each of those cards for what it actually is. It reads a delivery envelope, so a plan that lets a buyer carry more away lands higher on it, and that is a question about export permissions, marks and automation rather than about how a face holds together in motion. D-ID's own reading on that row is pulled down mostly by a mark its licence never lets go, which is a contract position and not anything visible in a frame.

Who the $29 D-ID Plan Fits, and Who Should Not Pay for It

D-ID at $29 a month on monthly billing suits a solo creator or a small team who wants conversational agents alongside flat talking-head video, for whom 1080p through the premium presenters is enough resolution, and who can live with a visible synthetic mark on the finished file. The rights position on that plan is clean, the same minute pool covers the API so automation costs no second subscription, and the licence permits the commercial use the cheaper plan does not.

Lite at $5.90 a month on monthly billing suits a different person entirely: someone whose videos stay inside a classroom, a family, an unmonetised hobby channel or a private demo. At that price the output carries a D-ID logo, the premium presenters are closed, and the agreement is explicit that the use is non-commercial. Paying for Lite and then publishing the result commercially puts a buyer outside the licence they agreed to, which is a worse outcome than paying more in the first place.

Nobody should buy any D-ID plan if the deliverable has to be unmarked. No tier removes the synthetic mark, the FAQ says so, the licence forbids removing it without written approval, and no amount of money changes that. If a client's brief rules out a generated-content mark, this product cannot meet it and a larger plan will not help.

Whoever buys, set the reminder before the money goes in. Fifteen days ahead of the billing date, not on the day the project ends, because section 8.5 makes the next charge unrecoverable and only the licence agreement carries the deadline.

And if the deciding question is whether these avatars will convince one particular audience, nothing on this card answers it. Every figure here describes what is paid, what may be done with the output, what the deliverable is, or what other paying customers reported and when.

The rules behind those four rows, including the things this desk will not claim to have done, are written up under how the scoring works, and the person who read the documents is named separately.