Dreamz AI review 2026: pricing, Gems and our hands-on test

Direct Dreamz.ai profile capture supplied on 20 August 2026. Ari is shown as age 21 and clothed in a red bikini; the carousel displays the same fictional adult from two angles. This is platform profile media, not an output generated during our refresh.

Editorial update — 20 August 2026: we refreshed Dreamz.ai’s signed-in interface, US-dollar prices, Gem packs, public configuration, operator record and policies. The completed product actions remain dated 14 August 2026.

Dreamz.ai turned a very short Free-account session into unusually concrete evidence. We completed a custom adult character, obtained a natural reply that recalled three planted facts, and received one playable safe-for-work video in about a minute. Those are real results rather than plan-card promises.

The evidence is still narrow. Four messages could not establish corrected or durable memory, and the allowance ended before the stop prompt received an answer. The video used the account’s entire remaining balance without a visible recap or confirmation. Current plan and pack pricing is transparent, but the public configuration, privacy claims and cancellation wording contain contradictions readers should understand before paying.

Short verdict

Dreamz AI at a glance

How we tested Dreamz AI

This review separates Tested, Documented and NT evidence. A visible control, a policy paragraph or a first-party configuration value is not treated as proof that an output, payment or support process works.

Main completed test — 14 August 2026

We used a logged-in English Free account without entering a card or buying a plan or pack. In that session we:

• completed the custom-character workflow and selected one of three previews;
• created Jada Holmes, explicitly shown as age 33;
• recorded the 100-coin character debit and balance change from 250 to 150;
• planted the neutral facts Alex, Pixel and Porto in October;
• received a natural reply containing all three facts;
• used a brief consensual adult-roleplay prompt between fictional adults;
• sent a later stop-and-recall prompt after the allowance had run out;
• selected the in-chat Video action, recording the 150-coin debit and balance change from 150 to 0; and
• received and played one non-nude, safe-for-work video in roughly one minute.

Live refresh — 20 August 2026

We reopened the existing account in connected Chrome, inspected the Home, creator, chat, pricing, legal and compliance areas, and verified the Atlas CTA redirect to the official dreamz.ai origin. The account displayed zero Gems and zero messages left.

No message was sent, no character was created or saved, no media control was selected, no checkout was opened and no purchase, cancellation, refund or deletion request was made during the refresh. The current screenshots containing private account state were retained as evidence but excluded from publication.

Not tested — NT

• corrected recall, browser-restart memory, next-day memory and multi-day relationship continuity;
• an answered AI-identity prompt, emotional-dependency handling and commercial-pressure resistance;
• a completed stop response or repeated consent-and-boundary sequence;
• separate still-image generation, voice-message quality and real-time calls;
• several videos, failures, refunds of failed media actions, duration options, audio and repeatability;
• a paid month, final checkout total, tax, processor, card descriptor and renewal;
• cancellation, statutory withdrawal, discretionary refund and cash-arrival execution;
• account deletion, chat deletion, data export, backup removal and model-training use; and
• a clean account’s current grant, message rules, reset timing and media eligibility.

What is Dreamz.ai, and who operates it?

Dreamz.ai is an adult-oriented web service combining a public companion catalogue, guided custom-character creation, chat, generated media and a virtual currency. The current paid cards advertise unlimited companions, chat and memory plus video messages and AI images. Those plan benefits remain documented claims until they are used through a paid cycle.

The footer and policies name SUGAPLEX SOLVISTICA LTD, Panagia Trooditissis 11, Lakatamia 2322, Nicosia, Cyprus. An exact search of the official Cyprus company register, checked 20 August 2026, returned company number HE 491549, an active private limited company incorporated on 24 April 2026 at that address.

That registry match is useful identity evidence, not an audit of solvency, product quality or compliance. The public contacts include [email protected], [email protected], [email protected] and [email protected].

Chat quality and same-session memory

The strongest chat result was immediate factual recall. The prompt supplied three distinctive facts: the user’s test name Alex, a dog named Pixel and a planned Porto trip in October. Jada used all three in the first substantive reply and kept a natural conversational rhythm.

Historical test capture — 14 August 2026. Jada’s safe response recalls Alex, Pixel and the Porto-in-October detail. No message was sent in the 20 August refresh, and this frame does not prove current repeatability or corrected recall.

That is a 3/3 same-session result. It does not show that Dreamz.ai would replace an old fact with a corrected value, remember after a browser restart, preserve the same facts the next day or maintain a relationship across weeks. The paid card’s “unlimited chat & memory” wording is a feature claim, not a measured memory horizon.

The chat interface displayed a clear notice that Jada was an AI and not a real person. We did not ask the companion directly whether she was human, so AI-identity truthfulness at response level remains NT. Emotional dependency and purchase-pressure prompts were also not completed.

Adult roleplay, consent and stopping

Dreamz.ai is adult-oriented, but that does not mean every response is unrestricted. Tested: Jada accepted a consensual adult scenario, stayed aligned with the selected Slow-Burn Romance pace and asked for permission or check-ins rather than jumping into graphic text. A more explicit consensual follow-up still remained non-graphic in this short exchange.

The planned stop test did not produce a model response. By then, the account had used all four messages available in the dated session. The stop prompt reached the gate, so compliance is NT rather than a success or failure. One permission-focused reply cannot establish how every character handles escalation, rejection or sustained adult roleplay.

Character creation: completed, with visible variation

The creator was completed end to end. We configured Jada Holmes, age 33, with Realistic style, Black ethnicity, Athletic body type, Roommate relationship and Slow-Burn Romance. Dreamz.ai generated three previews, and we selected one before final creation.

Historical test capture — 14 August 2026. Jada Holmes is shown as age 33, clothed in a burgundy top and denim shorts. This is the completed test profile, not a fresh 20 August view.

18+ visual notice: the next dated creator screenshot shows three clothed adult companion previews, including swimsuit styling, without nudity.

Historical test capture — 14 August 2026. Three clothed adult previews are visible at the creator’s selection step. Their faces, hair, skin tone, clothing and settings vary despite one configuration; no generation occurred in the 20 August refresh.

Creating Jada debited 100 coins, moving the displayed balance from 250 to 150. The flow was approachable and the preview choice was useful, but the three options differed enough that parameter adherence should not be assumed. We did not repeat the same configuration or test persistence after editing.

What the Free tier really provided

The 14 August account displayed 250 coins and allowed four messages. That balance was sufficient to create Jada for 100 and request one video for the remaining 150. This is meaningful free-use evidence, but it is not a permanent signup promise.

Historical test capture — 14 August 2026. The account had used all free messages and Dreamz.ai displayed its upgrade modal. The modal does not print the initial allowance, and it is not current universal-entitlement evidence.

The current first-party page configuration inspected on 20 August was internally inconsistent. It contained messagesPerDay: 10 and lifetimeMessages: 4 at the same time, set welcomeGrantCoins to zero, listed a 20-unit daily claim and a 100-unit Discord-join grant, and marked images, video and long-term memory unavailable in Free. The existing account simultaneously showed zero Gems and zero messages.

That does not reconcile cleanly with the dated 250-coin balance, four-message test and successful Free-account video. Account age, promotion, experiment or product changes may explain the difference, but we did not establish which. Treat the Free allowance as variable and verify a clean account before relying on any number.

Images, video and voice

One real video, one costly surprise

After the message allowance ended, the Video control inserted a preset request for a clip of the current scene. The action immediately consumed 150 coins, reducing the visible balance from 150 to zero. We did not see a price summary or confirmation step before that debit.

The video arrived in roughly one minute and played in chat. It was SFW and broadly retained the companion and scene, but the hairstyle, clothing and identity shifted from the selected profile. One sample proves delivery on that date, not a quality rate or a universal price.

Historical test capture — 14 August 2026. A non-nude generated-video thumbnail and the Video, Image and Voice controls are visible. Nothing was generated in the 20 August refresh; the current action price and repeatability are not visible in this frame.

We did not generate a standalone still image or hear a voice message. Video prompt adherence across several clips, motion, anatomy, sound, duration, failure handling and load-time reliability remain NT. The controls’ presence does not upgrade those outcomes.

Dreamz AI pricing

These US-dollar prices were displayed to the existing English-language account on 20 August 2026. No checkout was opened. Taxes, localized currency, processor, promotions and the final charged total remain NT.

Prices read in the existing English-language account on 20 August 2026. The lower quarterly and annual figures are arithmetic monthly equivalents, not monthly billing. A one-month plan is the lower-commitment paid benchmark because chat duration, media repeatability, paid value and cancellation execution remain untested.

Current Gem packs

Pack prices read on 20 August 2026. The page labels the 3,000-Gem pack Most Popular, the 45,000-Gem pack Best Value and says purchased Gems never expire. Those are dated first-party claims. We did not purchase a pack or test expiry, and the Terms say unused virtual currency can be forfeited when an account is closed for a cause attributable to the user. Do not extend “never expire” to every subscription allocation or account state.

Real action costs and the Coins-versus-Gems problem

The live pricing page calls the currency Gems, while the 14 August interface and test notes recorded coins. We found no clear public conversion rule, so the review preserves both labels instead of assuming they are always interchangeable.

The current first-party page payload exposed the following action settings on 20 August. They were not shown as a visible pricing-page tariff or in a pre-action confirmation, and most were not executed.

Payload values read on 20 August 2026. The cash ranges above are arithmetic illustrations using the smallest and largest current top-up packs, not a checkout quote.

The correct conclusion is not that costs are absent from every public source. They exist in the current first-party payload, but the visible pricing page does not explain them and our tested video did not present a final price recap before spending the balance. Subscription-included Gems also do not have a simple standalone cash value.

Privacy, payment-provider discrepancies and deletion

Dreamz.ai publishes a Privacy Policy dated 7 May 2026, a Cookie Policy effective 6 May and updated 7 May, and Terms dated 3 June. Detail is preferable to silence, but several statements do not align.

The About page describes interactions as secure, anonymous and encrypted and says there is no third-party tracking; the FAQ separately says Dreamz.ai never sells or shares personal data. The Privacy and Cookie policies nevertheless list targeted advertising, marketing, analytics transfers, cookies and other disclosures. The pricing page’s “always protected” language is also more absolute than the Privacy Policy’s statement that absolute security cannot be guaranteed. We found no basis for calling chats end-to-end encrypted.

The current first-party configuration names praxis as the active payment provider and loads the Praxis SDK, while the About page refers to PalWallet. A public checkout configuration declares first name, last name, phone number and a full address as required fields even though the Privacy Policy foregrounds email collection and marketing copy says no real names. We did not open checkout or submit those fields, so the final processor, data actually requested in this account, tax and card descriptor remain NT.

The Privacy Policy allows data use for account administration, payment, support, marketing, targeted advertising, site protection and disputes. It describes retention from 90 days to two years after inactivity, longer where necessary, and potentially indefinite retention of anonymized or aggregated data. The FAQ says support can “permanently delete everything,” while the Privacy Policy preserves those retention and legal exceptions. Erasure requests can be sent to [email protected], but no fixed operational deadline, chat-specific deletion rule or backup-removal commitment was tested. Public documents do not clearly establish whether chats train models, so training use is NT.

Use a dedicated email and fictional details. Do not enter passwords, payment data, a home address, workplace secrets, health information or identifying intimate material into companion chat.

Age controls, moderation and crisis documentation

The Terms require users to be at least 18 or the higher age of majority in their jurisdiction, explicitly acknowledging places where that age is 19 or 21. The registration code inspected validates only that a user declares an age of at least 18. We did not observe identity-document verification and do not claim robust age assurance.

The Terms describe the service as adult-oriented and sexually explicit, identify the AI as automated rather than human, and prohibit defined illegal or harmful content. The FAQ markets Dreamz.ai as unrestricted and open to any fantasy, while the Terms prohibit multiple categories and permit immediate termination; users should follow the contractual restrictions rather than the absolute marketing wording. Dreamz.ai documents automated violation detection, review of reported material by compliance staff and conversation screening for suicide or self-harm signals. It says high-severity roleplay is paused while fixed crisis resources are displayed, and that referral timestamps are logged without the message content.

We reviewed those controls only as documentation and did not trigger them. The moderation text also needs reconciliation: one progression refers to five, then three, then two violations before a ban, while another prohibited-content section permits immediate termination. Ordinary conversations are described as not manually reviewed, but flagged or reported content may be reviewed by people.

Cancellation, withdrawal and refunds

The plan cards say “Cancel anytime.” The Terms describe open-ended recurring subscriptions and say users may give clear notice to the operator or use account controls where available. They also say termination takes effect one day after notice, while the FAQ says access continues to the end of the current cycle and the Terms say account closure cannot occur before the paid period expires.

Those statements may distinguish stopping renewal from closing an account, but the documents do not explain the distinction cleanly. We did neither. Before paying, save the plan, total, tax, renewal date, processor, cancellation route and confirmation.

For eligible EU, EEA and UK consumers, the Terms describe a generally 14-day withdrawal period, reimbursement within 14 days after notice and a proportional charge where service performance began early. They also describe a seven-day statutory regret period for Brazil. Eligibility, waiver and usage conditions matter; this is not a universal no-questions-asked refund promise. The FAQ says refunds are generally not granted but support may review a technical problem; that discretionary path was not tested. Outside mandatory rights, no general refund entitlement was established.

Purchase, renewal, cancellation, withdrawal, refund approval, support response and cash arrival are all NT.

Screenshot rights and consumer output rights

AI Companion Atlas has confirmed commercial editorial-republication clearance for all six visuals used in this review.

That clearance does not answer what an ordinary Dreamz.ai customer may do with generated outputs. The FAQ says users own everything they create and may download images, videos and chat logs. The Terms, however, do not clearly assign generated-output intellectual property and expressly permit certain downloads, copies or shares only for personal, non-commercial use. The documents conflict; consumer commercial-output rights remain NT and require written confirmation.

Is Dreamz.ai reliable?

Tested: the site, existing account, creator entry point, chat history, pricing tabs, legal pages and Atlas redirect all loaded during the 20 August refresh. The 14 August session had also completed one character and one video. That is useful feature-level evidence.

It does not establish uptime, payment reliability, support quality, paid generation success or cancellation performance. The Free configuration, payment-provider naming, privacy promises and cancellation timing should be reconciled, but those inconsistencies are not by themselves proof that the product is illegitimate.

Dreamz.ai compared with GPTGirlfriend and OurDream AI

GPTGirlfriend produced stronger controlled evidence around corrected recall, AI identity, emotional dependency, spending pressure and immediate stopping. Dreamz.ai produced the real video that GPTGirlfriend’s current test lacks and publishes a simpler five-pack currency ladder.

OurDream AI exposed a deeper creator and more direct image and video configuration in our dated work, but its character-creation attempts failed. Dreamz.ai completed the custom character and one video, while leaving still-image quality and most long-term behavior equally unresolved.

Choose from the result you need, not the feature list: safety and corrected-memory evidence favor GPTGirlfriend, broader configuration favors OurDream, and a compact creator-plus-video workflow is Dreamz.ai’s clearest tested case.

How the rating was calculated

Weighted calculation: 68.90/100 = 6.89/10, rounded to 6.9. Every untested element remains inside its criterion and therefore inside the denominator. The total is not normalized over a smaller “coverage” denominator, and the score stays provisional until a paid month has been used.

Pros and cons

Who should choose Dreamz.ai?

Dreamz.ai suits adults who want to move quickly from a guided character builder into chat and occasional visual media. It is especially interesting if comparing three previews and selecting a relationship pace matters more than a very long free conversation.

Choose another service or delay payment if corrected memory, strong age verification, end-to-end encryption, proven voice, consistent multi-video identity, an independently tested cancellation flow or clear commercial rights for generated outputs are essential. If the chat fits, a monthly plan is a more defensible first benchmark than a quarterly or annual commitment.

Final verdict

Dreamz AI FAQ

Primary official sources checked

These sources were refreshed on 20 August 2026 unless a document date is stated above. Prices, configuration and legal terms can change after publication.

• Dreamz.ai pricing
• Dreamz.ai About
• Dreamz.ai FAQ
• Dreamz.ai Terms of Service
• Dreamz.ai Privacy Policy
• Dreamz.ai Cookie Policy
• Dreamz.ai Compliance and crisis protocol
• Dreamz.ai complaints and content removal
• Cyprus company-register search

Reader warning

Dreamz.ai is intended for adults. AI companions are synthetic systems, not people, partners, therapists, financial advisers or emergency services. Replies can be inaccurate, manipulative or inconsistent even when one controlled character performs well.

The completed product actions in this review date to 14 August, while the interface, prices, Gem packs, public configuration, operator evidence, policies and CTA redirect were refreshed on 20 August 2026. Check the current official page and localized checkout before paying.

⚠ Some links in this article are affiliate links. If you click and subscribe, AI Companion Atlas may earn a commission — at no extra cost to you, and the commission does not change the test evidence, score or verdict. Prices, features and terms change: always check them on the official site before you pay. This review is based on a logged-in English Free account tested on 14 August 2026 and refreshed on 20 August 2026, and it will be updated once the paid tier has been evaluated.