Hyper3D Rodin

By Minel Gunesoglu, founder of Vouch. The work behind this page was documentary: the vendor's terms, fulfillment policy and pricing page read in both billing views on one day, fifty-three Trustpilot entries read across three pages including the company's own replies, and a review count corrected twice in our own records before the usability cell was defensible. Every claim below about what this tool produces is attributed to somebody else's dated record, because this desk runs no generation batteries of its own. Published 14 August 2026, re-checked monthly.

Search "Rodin" on its own and the first page of results belongs to the French sculptor, not to any software. Add one word and a second hazard opens: the App Store carries an unrelated app called "Rodin ai - 3D Model Generator", published by a developer named Mahmood Khan, rated 1.0 out of 5 on two ratings. The product measured here is Hyper3D Rodin, sold at hyper3d.ai by Deemos Corporation of Wilmington, Delaware, and most third-party coverage titles it Rodin AI. Every rating and figure below belongs to that product and to nothing else wearing the name.

Past the name, the public record splits along a line that matters more than the score. Wherever people discuss what Hyper3D Rodin makes, the verdict is narrow and warm: clean meshes, usable topology, a part-splitting feature they name by its product name. Wherever people discuss being billed, and that is one platform only, the record turns hard. Trustpilot's topic row for this company opens with Subscription and Cancellation, and across nine months of dated reviews the most repeated specific complaint is that the cancel control did not work. On 3 August 2026 Deemos replied to one of those reviews and named the cause itself: "A database issue occurred a few hours ago, which affected the cancellation process for some users." A complaint is one person's account. A complaint beside the company's own dated concession is something else.

What Hyper3D Rodin Is, and Why Searching Rodin Alone Won't Find It

Hyper3D Rodin turns a written prompt or an uploaded photograph into a textured 3D mesh with an editable topology, and it is one of three products Deemos runs on a single account and a single credit pool. The other two are ChatAvatar, for talking-head avatars and 3D previews, and OmniCraft, a toolkit of image-remix, texture, HDRI and mesh-editing utilities. Only the first is scored here, and the shared account matters twice over: credits bought for one product are spendable on another, and the public reputation record covers the whole company rather than the 3D generator alone.

The collision runs past the sculptor. A Microsoft Research 3D-avatar project carries the same name, as did a VR authoring tool from 2017, and software directories list several products differing from this one by a character or two. The vendor's compound, Hyper3D Rodin, is the only unambiguous string; "Rodin AI" is what the wider web uses and works in practice, since those two extra letters drop the sculptor out of the results entirely.

The App Store App Named Rodin ai, Rated 1.0/5, Is Not Hyper3D Rodin

One collision is worth isolating because it can do direct financial damage to a reader comparing on a phone. The App Store listing "Rodin ai - 3D Model Generator" names Mahmood Khan as its seller, carries a 1.0 out of 5 rating from two ratings. One more field is worth putting on the record because it is easy to miss and it comes from Apple's own data rather than from anyone's reading of it: the bundle identifier Apple stores for that app is com.rodin.ai.hyper3d.ai.rodinAI, which contains the vendor's domain as a string. What that string means is not something this page can tell you, since a bundle identifier is a value a developer types and nothing here supports a claim about why it was typed. It is recorded because a reader checking whether two listings are the same product will find it, and better to meet it with its date attached. Read on Apple's public app search endpoint, 14 August 2026. Nothing about that app's rating or reviews is evidence about the tool reviewed here, and any write-up quoting a one-star App Store average for Rodin AI has almost certainly imported it from the wrong product.

Hyper3D Rodin Reviews on Trustpilot: What Nine Months of Cancellation Reports Show

Trustpilot's profile for hyper3d.ai is the only consumer-review platform with a usable sample for this company, and it is worth naming what that means rather than leaving it as a claim. G2 was tried on 14 August 2026 under both of the slugs this product appears under, and every method returned a block rather than a page, including the one that got through to Trustpilot. Capterra carries no listing for it at all. So this section rests on one platform not by preference but because the others gave us nothing to read. Read through a regional mirror after the main domain refused every direct request, the profile showed 3.1 out of 5 across the 53 reviews it counts. Forty-seven of them parsed as entries in the capture, and those span 29 November 2025 to 13 August 2026. Two scope caveats travel with that figure everywhere it appears below. The profile rates Deemos as a company, so ChatAvatar and OmniCraft experiences sit in the same average as ones about the 3D generator. And 53 is thin for a consumer aggregate, which is why the usability cell publishes a figure pulled toward the middle rather than the raw conversion.

What the sample lacks in size the platform makes up in labelling, which spares this page from characterising anything. Trustpilot's own top-mentions row runs Subscription, Cancellation, Customer communications, Quality, Service, Refund. Its automated summary, marked "Created with AI, based on recent reviews" and scoped to 48 of them, reaches the same split: reviewers "generally find the platform very easy to use", and separately "Some reviewers were not happy with the subscription management, noting that cancellation options frequently fail to work properly across the platform's website."

Hyper3D Rodin's Cancel-Button Complaints, and Deemos's Own 3 August 2026 Reply

The recurring complaint describes a control rather than a policy. The oldest review read here, dated 29 November 2025 and titled "Cannot cancel trial, was charged $300", reports the in-product command failing and, at the time of writing, no reply to a support ticket: "I sent an email with screenshot and video-capture, no answer either. So I was charged $300 from my account." Deemos did answer, publicly, beneath that review: "We're truly sorry for this experience. That's on us. We'll check your case immediately and process your refund." Whether the refund followed is not something the record shows. Reviews describing the same failing control appear again in January, June, July and August 2026.

Then the 3 August 2026 pairing, the strongest single piece of evidence on this page. A reviewer wrote: "I can't cancel; the support button doesn't work, and the button to cancel the plan doesn't appear." Deemos replied the same day: "A database issue occurred a few hours ago, which affected the cancellation process for some users. The issue has now been fully resolved, and the cancellation function should be working normally again." One caveat belongs with that quotation: the review it answers carries a Portuguese title and Trustpilot displayed the exchange in English, so the reply text may be the platform's translation rather than the company's original wording.

Deemos disputes these accounts, and its rebuttal belongs beside them rather than under them. Answering a 26 July 2026 review the company contested that reviewer's facts directly, not just the conclusion drawn from them: "the claims in this review are not accurate. The user contacted our team through Discord before posting this review, and the subscription had already been successfully canceled by the time the review was submitted." It went on: "Our cancellation process is clearly available to all users, and we do not prevent, delay, or restrict subscription cancellations. We have also tested both the cancellation flow and our AI support bot, and they are currently working as expected." Other replies point to a named route through a support agent in the corner of the website. Reading both sides together, the dated record supports something narrower than either: a documented cancellation path exists, a defect in it was acknowledged by the vendor on at least one occasion, and reports of the control failing recur across nine months rather than clustering on one day.

Hyper3D Rodin's $24 Trial, the $120 Renewal, and a Plan You Cannot Cancel Mid-Term

A second cluster is about money rather than the button. The most quoted example is dated 30 May 2026, under the title "AVOID huge scammers": "the trial is $24 but then jumps up to $120/month if you dont cancel". Deemos answered with a pricing clarification rather than a denial: "The standard price of the Business plan is $120 per month. We are currently offering promotional discounts, which is why the first month may be priced lower." A reply on 24 June 2026 names a different entry price again, telling that reviewer "you initially subscribed to the $30 Creator plan through our $1 first-month promotion".

Set those beside the contract and the complaint stops being a matter of opinion. Deemos's Fulfillment Policy, read 14 August 2026, states: "Once a Subscription Plan is purchased, it cannot be cancelled or downgraded during the active subscription period." The Terms of Service repeat it at section 4.2, and section 4.1 adds a constraint the pricing page never mentions: "Users are required to pay for at least a two-month Subscription Plan in one lump sum." A discounted first month is therefore not a month you can walk away from at will, on the vendor's own documents. Upgrades are permitted mid-period; downgrades and cancellations are not.

Hyper3D Rodin's Pay-by-Result Credits: Praised in One Review, a $300 Charge in Another

The same billing mechanism produces the warmest and the coldest reviews on the profile. On 28 December 2025 a reviewer praising the output added: "I'm suprised that they won't charge anything until you're satisfied with the results." On 5 August 2026 another described the opposite experience of the same system: "Also you take TOKENS just for DOWNLOAD and don't say even a word about it until I see my zero balance just because of downloads".

Deemos's reply to that second review draws a line worth quoting exactly, because it contradicts the shorthand nearly every write-up about this tool uses: "downloading your own generated models does not require credits. Credits are only used for actions such as model confirmation, high-accuracy generation, or downloading other users' models." The mechanic is pay-on-confirmation, not pay-per-download. Whether an account experienced it that way is a separate question the record answers both ways, and one 5 July 2026 reviewer who reported credits clawed back with no refund later appended an edit: "EDIT: They finally refunded me."

What Hyper3D Rodin Users Report About Output Quality: Clean Topology on Two Unconnected Platforms

Two populations that do not read each other land on the same narrow claim. On Trustpilot, 28 December 2025: "Clean meshes and sharp details. Much better topology than I expected." On 3 February 2026: "The geometry quality is good enough that I can focus on design decisions instead of cleanup."

Separately, on the blind-vote arena at top3d.ai, fifteen community reviews dated between 1 November 2025 and 9 May 2026 converge on mesh cleanliness and part splitting, with a recurring reservation about PBR texture. From 12 March 2026: "The mesh quality is good and exports work smoothly. Sometimes the textures come out a bit flat, but for quick prototyping it's really useful." And from 1 December 2025, the confirmation mechanic again on a different platform: "You generate first and only pay credits after confirming, which saves a lot!"

The sharpest illustration of the split arrived on 13 August 2026 in one review. Its author reported being charged for a year after a trial without an email beforehand, wrote that the missed cancellation was "My mistake", recorded that "Happily they have begun the refund process (I was within 6 days)", and then closed with: "Produced the best mesh results from my testing to date". Deemos replied beneath it that "the refund has already been processed for you", and asked the reviewer to consider removing the review. One customer, a billing complaint, a refund the company confirms it completed, and unqualified praise for the output, in one entry.

How Hyper3D Rodin's 63.8/100 Score Breaks Down Across Four Dimensions

Four cells make this card, and they are four different kinds of thing rather than four readings of one thing. One converts a vote count, one a consumer rating, one a price list, one a contract. Only the first two sample anybody's experience, and only the first is deep.

Composite: 63.8 out of 100, which is 3.2 on the five-point scale, and the band chip beside it grades that number rather than the strength of the evidence beneath it. The composite is the geometric mean of the four cells, so a low cell drags harder than a high cell lifts, and 54.0 is the one dragging. One thing in the row above is worth separating before somebody quotes it back: the 57.7 per cent beside the capability figure is a win rate from the arena's closed season (how often this model's output was picked in a blind pairing) and it is not a score, not a percentage of anything this card measures, and not related to the composite.

The usability cell needs a warning label the other three do not. An axis called usability normally measures friction in an interface. Here it converts a record whose dominant subjects, by Trustpilot's own extraction, are subscription and cancellation. The same axis name on another card in this category may rest on an app-store or product-directory sample, where reviewers describe the software rather than the checkout. Same label, different underlying record, so a cell here and a cell elsewhere are not two readings of one scale. What goes into a score card here, and what deliberately stays out of it, is the same for every tool measured on this site; what feeds a given axis is not.

Hyper3D Rodin's 66.9 Capability Score: Rank 3 of 17 in a Closed Arena Season, and the Live Board's Lower Number

Capability converts one row on one public board. top3d.ai shows a visitor two generations grown from the same prompt with both makers concealed, takes the vote, and accumulates the verdicts into an ELO figure. This card reads a closed season: across "Jul 26", which ran from 1 July to 1 August 2026 and cannot move again, the row "Rodin Gen 2.5" finished third of seventeen at ELO 1077 on 1,388 votes, 801 won against 587 lost, with every entry in the field clearing a thousand votes. What that measures is which render a stranger preferred between two pictures, which is a real signal and a narrow one.

Reading the live board on 14 August 2026 shows why the closed season was chosen. That day the per-tool page published Overall ELO 1047, Geometry ELO 1024 and 412 votes at a 57 per cent win rate: rank unchanged at third, rating thirty points below the scored figure on under a third of the votes, and different metric labels, since the live page publishes an Overall and a Geometry ELO where the season table publishes a Textured track. A figure copied forward from a board that keeps voting quietly stops being true, which is what a closed season prevents. The same page also carries visitor ratings on a separate five-point scale, 40 of them, running HD Mesh 3.9, Low-Poly Mesh 3.9, Texture 3.8, Texture PBR 3.7 and Parts 4.0. Those are opinions rather than blind pairings, they sit lowest exactly where the community reviews above are most reserved, and none of them feeds the card.

Two dated sources also answer the speed question differently, and both ship because neither can be checked against the other. Deemos publishes a testimonial on its own pricing page reading: "Rodin Gen-2.5: Geometry in ~4s, full model in ~5s, 10M+ polygons, clean structure, production-ready outputs." top3d.ai's per-tool page, read the same day, states: "Rodin Gen 2.5 reports an average generation cost of $0.29 per test with a typical runtime of 4.5 min." A vendor-hosted testimonial and a third-party arena are not measuring the same thing under the same conditions, and nothing in either source explains the gap.

Is Hyper3D Rodin Good for 3D Printing? Bang to Parts, and What Is Not Settled

The one printing-specific feature with real corroboration is a named part-splitter, and it surfaces unprompted on both platforms read for this page. A Trustpilot reviewer on 29 January 2026: "For 3D printing, Bang to Parts is a huge improvement over manual splitting. It still requires some planning, but it saves a lot of repetitive work." The arena listing carries "Split into Parts" among its feature tags, its visitor ratings score Parts highest of the five facets, and a community reviewer there on 1 November 2025 wrote simply: "Very good split into parts feature". Splitting a model into printable parts is normally manual work, so that convergence is worth something.

Everything past it is unsettled. No export format list appears on the pricing page or the arena listing captured for this review, so no STL or watertight claim is reported here in either direction, and no independent print run against this tool has been published anywhere reached. Deemos markets a 3D Printing use case on its own site; that is marketing, not a measurement, and a page refusing vendor benchmarks in its score card does not get to borrow one four sections later. Print a cheap test object before committing a long one, and treat any printability figure quoted elsewhere as unverified until you can open its source.

Hyper3D Rodin's Commercial Terms: One-Way Indemnity, a $100 Liability Cap, and No Stated Free-Tier Licence

The fourth cell reads clause positions off documents anyone can open, which is the only cell here that does not depend on a platform serving a page that day. Four positions were looked for and three were answered.

Indemnity runs one way. Section 9 of the Terms of Service reads: "The Users accept to defend, indemnify, and hold harmless us and our subsidiaries and affiliates". No clause runs the other direction, so a user whose generated model draws a third-party claim defends it alone. That direction is the norm across this category rather than a quirk of this vendor, and it is reported because it answers the exact question a commercial buyer should ask, not because it is unusual.

Liability is capped, and the figure is unusually concrete. Section 10 states that Deemos's maximum aggregate liability will not "EXCEED THE GREATER OF $100 OR THE AMOUNT PAID BY YOU TO USE OUR SERVICES." Section 21 sends disputes to individual arbitration with a class-action waiver under Delaware law, and provides that if the waiver is held unenforceable as to a claim, that claim is severed and brought into the courts of Hong Kong. Those are the document's terms as written on 14 August 2026, quoted rather than interpreted.

Output rights are the sub-reading that lifts this cell. Section 5(b) is short and broad: "If you use Rodin to generate Output, we will not limit your use of such Output, subject to any restrictions set forth in these Agreements". Two things about that sentence matter. It carries no split between free and paid accounts, while section 5(a) does exactly that for ChatAvatar, placing unpurchased avatar output under a Creative Commons licence that bars commercial use. And it grants nothing about ownership: the same section states the company makes no representation that any output can be registered or credited to anyone. So the free tier's commercial position is not stated at all in the clause covering this generator, which is why it is recorded as absent rather than scored in either direction.

Refunds are the fourth position and the bluntest. Both documents agree, the Terms in capitals at 4.4 and the Fulfillment Policy in plain text: "All payments made on our platform are final and non-refundable, except as required by applicable law." Purchased credits are non-redeemable and non-transferable, and invoices must be disputed within seven days. Read beside the cancellation reports above, the contract does not contradict the complaints; it is why they take the shape they do.

What Hyper3D Rodin Costs: Pricing by Model Count, Not by Credit

Deemos publishes plan output in models rather than credit allowances, which is unusual and, for once, more useful. The pricing page, read 14 August 2026 in both billing states, prints an "Estimated output" line on each paid tier, so dividing it into the price gives a cost per model directly. Monthly is the basis published below, matching the basis the other cards in this category rest on, with the annual equivalent stated beside it rather than instead of it. No plan credit allowance appears anywhere on that page, so any per-tier credit figure quoted elsewhere for this tool cannot be reproduced from the vendor's own page, and none is printed here.

Every figure there was read from hyper3d.ai/pricing on 14 August 2026, and that date is load-bearing rather than decorative: a tier can be repriced between one reading and the next, and nothing on this page would know. The "Save 20%" badge checks out on both tiers, which is not true of every vendor in this corpus: $24 against $30 and $96 against $120 are each exactly a fifth off, and the annual totals divide back cleanly. The two cost-per-model figures are derived from the vendor's own numbers and inherit the vendor's own hedge, the word "About". What they show is that four times the price buys roughly seven times the output, so the step up is cheaper per model than it looks. Top-ups sit outside that arithmetic entirely, at a stated "Direct credits cost $1.5 per credit." One basis warning, because a third-party listing already gets it wrong: top3d.ai advertises this tool "From US$24.00/mo", which is the annual-billed equivalent rather than a monthly price, and anyone budgeting from that line will be a fifth short.

Hyper3D Rodin's Free Tier: Pay by Result, Not a Monthly Credit Cap

The free plan works on a different principle from a monthly generation allowance, and the difference decides who it suits. Its headline on the pricing page reads "Start creating and pay only when you confirm results.", with "Generate for free before confirmation" among the listed features. The pricing FAQ states the rule for every tier: "Generation is pay by result: you can preview results before confirming credit spend."

Two limits sit on that plan and neither is a generation count: export is restricted to legacy models, with unlimited export and any use beginning at Creator, and the workspace holds ten private assets. A free account can therefore iterate as much as it likes and find out what the tool produces; what it cannot do is take current-generation output away for work. Reviewers experience that as generous or as a wall depending on which side of confirmation they hit, which is the disagreement quoted earlier.

Hyper3D Rodin Pricing Tiers: About $0.50 a Model on Creator, $0.29 on Business

Creator at $30 a month is aimed at one person producing assets regularly, and its listed limits are the ones a working user meets: multi-image input, smart low-poly, HD and custom texture, and a redo budget of twenty geometry and six material retries. Business at $120 adds 4K textures, high-poly quads, fifty geometry and fifteen material redos, and full API access. Enterprise is quoted rather than priced and covers on-premise deployment and custom fine-tuning.

One line on the Business tier reads like a 3D feature and is not. Its "Image & Video" block resells third-party generation models by name, listing Nano Banana 2, Veo 3.1 and Seedance X30, where Creator's equivalent lists Seedream and default models. That bundling is a menu fact, and it explains a Trustpilot review from 13 August 2026 whose entire complaint concerns a Veo 3.1 video feature failing rather than anything 3D; how those bundled models perform belongs to their own records, not to this card. One arithmetic coincidence also deserves a flat sentence: the Business cost per model computed from the vendor's page, about $0.29, is the same number top3d.ai publishes as this tool's average generation cost, neither source explains the other, and no relationship between them is asserted here.

Hyper3D Rodin's Education Tier: A Price Published Only After Verification

Students and educators get a tier with no number on it. The pricing page shows an Education card carrying a description and a "Verify education" button and no figure of any kind, and the FAQ says only that "Education includes Creator tier benefits at education pricing, but it requires verification for eligible students, educators, classrooms, or learning programs." At education pricing is the whole disclosure. A tier priced only after an eligibility check cannot be compared against anything before you apply, so treat any specific education figure quoted for this tool elsewhere as unsourced until the vendor publishes one.

Hyper3D Rodin's API: What a Business-Tier Key Costs, and the Rate Limits It Carries

For a developer the answer is a price rather than a feature: full API access appears in the Business tier and in neither Free nor Creator, so a key costs $120 a month, or $96 a month billed annually.

Throughput is documented but the documentation is ambiguous, which is worth knowing before a pipeline is planned around it. The Business card states "120 RPM on first three tiers; 240 RPM on higher tiers", and the plan grid repeats it as "120 or 240 RPM" in the Business column. Nothing on the page defines those tiers, and they cannot be the subscription plans, since neither Free nor Creator carries API access at all. Anyone sizing a batch job should confirm the applicable rate limit with Deemos first, because a request ceiling nobody can locate in the documentation is a ceiling you meet in production. A third-party model marketplace also lists this generator for API use; that listing was found in a search on 14 August 2026 and not verified against a saved capture, so nothing about its rates is reported here.

Alternatives to Hyper3D Rodin: Meshy and Tripo AI, Named Without a Ranking

Meshy and Tripo AI carry dated cards here as of 14 August 2026, and this section prints neither composite. Each of those cards was assembled from whatever evidence its vendor and its review platforms happened to leave behind, and on this page that is not an abstract worry. The usability figure here is built from a record where the dominant subjects are subscription and cancellation; the equivalent cells on the neighbouring cards come from platforms where people are talking about using the product. Two numbers gathered that differently answer the same question only by accident, so what follows names what each tool does and links to the card that shows its working.

What can be said is what each free tier does, since that is the decision most readers arrive with. Meshy renders output from its current flagship model on a free account and then withholds the file, a restriction documented in the vendor's help centre rather than on its pricing page. Tripo AI lets a free account export but meters it, at 200 credits a month against a separate download ceiling, read 10 August 2026, with free downloads restricted to an older model version. Hyper3D Rodin does neither: generation and preview are unmetered, the charge lands at confirmation, and export is restricted to legacy models. Three answers to one question, none of which says anything about which mesh is better.

One caution about wherever you read the next comparison, this page included. Some of the most visible write-ups about this product sit on the vendor's own blog, and some of the most visible write-ups about rival tools are published by their competitors. Ask who profits when a particular tool comes out on top, and treat an undated figure as no figure. For the wider view, the 3D model generators measured on this site are listed with their cards.

Hyper3D Rodin Verdict: Who the Free Pay-by-Result Tier Suits, and Who Should Not Subscribe

Start on the free tier and understand what it is for. Generation and preview cost nothing, which makes it a genuinely good way to find out whether this tool produces what you want before any money moves. Current-generation output will not leave that account, so treat it as an evaluation budget and nothing more.

Pay the $30 if you are a solo creator or a small team producing concept geometry, prototype props or printable parts, you expect to finish the mesh yourself, and part splitting is worth something to you, because that is the one capability with independent corroboration on two unconnected platforms. Pay the $120 only if you need the API, since that is what the step actually gates, and note that on the vendor's own model counts the larger plan is the cheaper one per model. Before paying either, read section 4.1 of the Terms and decide whether a two-month minimum commitment is one you want, because a discounted first month is not a month you can leave.

Do not subscribe if your decision assumes you can cancel on a whim. The Fulfillment Policy says a plan cannot be cancelled or downgraded inside its period, payments are final and non-refundable except where law requires otherwise, and nine months of dated reports describe the in-product control failing, one of which the company itself answered by naming a database issue. Deemos disputes the wider characterisation and points to a working route through its support agent; both readings are above, with dates. Do not subscribe if your own client contract obliges you to pass along protection you do not hold. Section 9 runs in one direction only, and section 10 puts a ceiling of $100 or the sum already paid, whichever is greater, on what Deemos can owe. Read both before an asset generated here goes into work somebody else signs off on. And do not buy on a rating you found quoted rather than opened, since one of the names closest to this product on the App Store belongs to somebody else entirely.

A composite of 63.8 is silent on the question that decides most 3D purchases: what happens after the download, when the file meets a rig, a UV unwrap or a slicer. Answering that would mean generating a model and opening it, and no cell on this card was built that way. The reproducible part of the picture is printed above, with its dates and its links attached.

Score card built and page written by Minel Gunesoglu, founder of Vouch. LinkedIn. Hyper3D Rodin's pricing, terms, fulfillment policy and Trustpilot profile were each read on 14 August 2026; the arena season behind the capability cell was read on 2 August 2026 and is closed, so it does not move. Disclosure: no affiliate relationship exists with Deemos or Hyper3D, so every vendor link above is a plain link; where affiliate links appear elsewhere on this site they are marked, and a paid submission buys review speed only, never a listing and never a score.