Nomi AI Review 2026: Price the Memory Habit, Not the First Chat

Disclosure: AISoul publishes this review and sells a different adult companion product. The linked AI Girlfriend Lab review is previously supplied commercial evidence, not a fresh independent validation of the current release or a new test performed for this update. October 5, 2026 checks below are public-document checks; no fresh paid Nomi account, controlled benchmark or private checkout was used.

Is Nomi the right fit for your companion habit?

Nomi is a documented-feature fit for someone who wants conversation continuity plus optional media and voice. It is not automatically the best choice for every photo, character-authoring or privacy requirement. Set the must-have task before judging a warm first response.

Concentrating your trial on one relationship makes the results easier to interpret. It is a recommendation for the trial, not a claim that Nomi only supports one character; its documentation also covers multiple Nomis and group chats.

Treat continuity as Nomi's primary product

Nomi positions continuity and evolving relationships as central features. Those vendor descriptions are not measured recall rates. A useful observation is whether a companion returns a harmless fact after a gap and then applies a later correction without you putting the answer into the question.

Keep three things separate: remembered facts, relationship tone and visual identity. A correct coffee preference does not prove the tone suits you, and a good selfie does not prove long-term recall. Nor can one wrong reply reveal whether the hidden cause was storage, retrieval, instructions, model changes or ordinary variation. Test the behavior you need instead of treating the word “memory” as a quality score.

Read the 21-day memory probe within its boundary

The cited commercial AI Girlfriend Lab Nomi review labels its test July 24, 2026, reports a July 19 test end and shows an August 27 page update. Its account describes web and mobile web, trial and paid use, with three evolving facts checked at day 7 and day 21. It reports 3/3 at both checkpoints, a calm text-first experience and perceived warmth from voice notes.

That is the source's bounded observation, not a result reproduced here. The public record does not justify this article's former extra claims about a fresh account, desktop Chrome or the absence of model changes throughout the window; those claims have been removed. It also contains conflicting overall display scores, so this review does not adopt them. A denominator of three facts in that test cannot be converted into a universal memory-success rate, a cross-product win or a promise about another user's conversation.

Use the observation as a reason to try a return test, not to skip one. Keep the actual plan, thread, dates, prompts, intervening message count and correction wording with your own result. Public feature documentation establishes available controls, not whether your next reply will use them correctly.

What can you edit in Nomi's memory?

Nomi exposes Shared Notes for stable context and Mind Map summaries as an overview, not a complete editable dump of every underlying memory. Mind Map 2.0 describes graph and table views with editing and adding information. It is a higher-level layer that develops as conversations accumulate, so a new account without a developed map is not evidence of no memory.

The current correction guide makes an important distinction: editing a Mind Map entry does not change or delete the underlying memories. A summary can be edited and locked against automatic updates, but that does not guarantee a conflicting old fact disappears from later replies.

For a harmless correction, record the wrong fact, state the replacement clearly in chat, and update the relevant exposed note/summary where supported. Ask a neutral question later without embedding the answer; check both the displayed entry and the reply. If the conflict returns, record that result and consult the correction guide/support rather than repeatedly resetting the whole relationship. Do not use summary editing as an account-data deletion method.

How do you start a low-risk Nomi trial?

Start with a fictional or non-sensitive setup and one thread you can revisit. Nomi 101 documents choosing relationship, personality, interests and appearance, using Shared Notes, and clarifying behavior with an OOC instruction. You do not have to manufacture an elaborate backstory to test the basic workflow.

1. Nomi's current privacy policy describes Google/Apple account login followed by a name or pseudonym and date of birth. Read the actual eligibility/data prompts and free allowance first; do not assume a universal extra identity-document gate from an old review.

2. Choose one Nomi and relationship style. Add only the stable fictional details you actually need in Shared Notes.

3. Have an ordinary conversation and enter three harmless facts. Save the exact wording and a neutral later cue; do not put the answer inside the test question.

4. Return to the same thread after a gap. Record elapsed time and the number of intervening messages, then test one correction separately.

5. Try a permitted selfie or the desired voice/media route before buying. Record the shown charge or plan wall instead of assuming a feature name means free access.

If tone or roleplay drifts, a clear OOC instruction is a documented clarification route, not proof you can force perfect recall. Set a stop condition before paying—for example, unresolved factual conflict after the same correction check—not after an attractive first chat.

Selfies, uploaded pictures, generated videos and voice are different features

Nomi's selfie guide describes pressing the camera button, continuing to chat while the image is prepared, and receiving the ready selfie in the conversation and album. Recent context and selfie settings can influence it. The cited guide describes a nudity filter and automatic refund of failed selfie requests; do not infer unrestricted image permission from adult text. Confirm the active image system's current rules.

For current visual controls, the V5 guide, updated August 20, documents selecting V5 under the Nomi's Image Settings, choosing a stock Anchor and adjusting fidelity. Selfies follow conversation context; art supports more direct written prompts. Stock Anchors let you start without creating a custom one, and the guide acknowledges possible appearance drift. This is a control description, not a measured same-face score. The August 23 nude-photo FAQ says nude selfies are not permitted; flirty clothed images are different. Do not use an older stream discussion of more permissive adult content to promise nude access.

Sending a picture and understanding one you upload are separate workflows. Nomi 101 documents an upload control for image understanding; the selfie guide says you cannot upload your own image to create a Nomi, to prevent misuse such as deepfakes. That creation restriction does not mean all image uploads are unavailable. Do not use another person's private photographs without permission.

The video guide documents starting from an enlarged album image or the art menu, entering a motion prompt and viewing the result in the album. This is image-to-video generation, not a live camera or two-way video call. The guide's duration and processing-time estimates are dated; this review does not claim a fresh video-speed measurement.

The current credit-values wiki, updated August 24, lists 1 credit for a selfie or art request, 6 for a standard-definition video request and 10 for a high-definition video request. Purchased credits do not expire; buying them requires an active subscription. For tools with a replenishing daily limit, credits are used after that limit is reached. These units do not establish the USD pack price or your remaining daily video allowance—check both in the account before submitting.

For voice, the current voice wiki documents paid voice messages/calls in the same conversation context. It distinguishes unlimited in-house/custom-voice use from connecting your own ElevenLabs account, whose quota and charges depend on that separate service. Unlimited voice does not establish unlimited images, videos or unrestricted content.

What are Nomi's published free limits and paid prices?

Nomi's refund FAQ, updated August 23 and reopened October 5, describes an indefinite free plan with 50 messages per day. Its March 20 vendor comparison, also reopened October 5, publishes 2 free images/day and 40 subscriber image credits/day. These are published rules, not verified balances in a new account or a promise about every media tool.

Those values appear both in the July 24 Lab capture and the March 20 official publication reopened for this update. Reopening an older publication is not a fresh personalized checkout: regional/store charges, taxes, promotions and account eligibility remain to be checked. The equivalent monthly column is arithmetic, not a monthly installment offer.

Paid access recurs. The official policies generally treat subscription purchases as nonrefundable; unused time or disappointment is not a reason to assume a refund. Start with the free workflow before making an annual commitment, and copy the actual renewal terms from the purchase channel.

How do you cancel Nomi without confusing cancellation and deletion?

Cancel through the channel that billed you, before its renewal date. Nomi's cancellation policy distinguishes web, Apple and Google subscriptions and says cancellation preserves the remaining paid period. Deleting the app is not cancellation.

- Web purchase: Profile → Subscription → Manage Subscription → Cancel Subscription; complete the confirmation.

- Apple purchase: device Settings → Apple ID → Subscriptions → Nomi → cancel.

- Google Play purchase: Play Store → Payments & subscriptions → Subscriptions → Nomi → cancel.

Record the confirmation and paid-access end date. If you also want to delete the account, treat that as a separate data-removal request; do not assume an account or app deletion has canceled a store subscription. These are policy instructions, not a cancellation we performed in a live paid account.

What does Nomi's privacy policy actually promise?

Nomi's privacy policy, updated April 27, 2026, identifies Glimpse as operator and describes collecting account information, chats/customization, payment-related data and usage/technical information. Its promise not to sell or rent information does not mean nobody processes the conversation or that all data stays on your device.

The policy describes account deletion through Settings, generally taking about 28 days, with exceptions for legal obligations, communications and certain archives. It also describes training archives retained in a form no longer attributable to the deleted person. Do not translate that into immediate erasure of every copy. Neither a Mind Map edit nor deleting a visible message establishes account-wide deletion.

Keep identifying documents, passwords and sensitive personal facts out of a trial. This review makes no end-to-end-encryption, absolute-security, therapy or emotional-health benefit claim. If a particular retention or processing condition is essential, verify it with the operator before sharing—not after testing whether the companion remembers it.

Complete the Memory-Habit Value Ledger

The ledger forces habit and cost onto the same page instead of turning one successful callback into a buying recommendation.

Hypothetical routing example: a daily user may continue only if two return checks meet a self-chosen recall threshold and the journal effort remains acceptable. A visual shopper may reject Nomi before payment if the live media route cannot meet a required image workflow. These are conditional routes, not fabricated outcomes. The July 3/3 result belongs to the published test alone and is not prefilled as the reader's result.

Separate remembered facts from relationship judgment

Memory of discrete facts can be checked after a gap. Relationship judgment—tone consistency, emotional arc and roleplay fit—is a separate evaluation. A companion can recall a favorite place yet respond in an unwanted register, so the ledger records recall and correction burden independently. Ask about each planted fact without embedding the answer, then record whether it appears and whether a later correction persists. When the result misses the threshold set before the trial, the observed workflow has failed that user's requirement even if the vendor page still says “long-term memory.” This does not diagnose the hidden cause.

Check current price, identity gates and media before paying

Before entering payment details, record the account's actual first charge, recurring charge, interval, cancellation channel and media allowance. The public rules above give a starting point, but do not establish your region's checkout, every identity-verification trigger or every video's price. Treat an unavailable required media route as a reason to stop, not an invitation to guess the premium entitlement. Memory features do not establish privacy or deletion behavior.

Choose Nomi only when you will maintain the memory loop

Set the intended weekly habit, recall threshold, acceptable setup time and media requirement before the trial. Nomi is worth considering when the live product meets those requirements and you expect to use the continuity workflow—not simply because one greeting feels warm. Stop or cancel if the actual cost or correction effort exceeds your preset boundary.

If the issue is explicit cue-based journals and editable context, compare Kindroid using AI girlfriend apps with memory. If you prefer visible personal facts, compare Nomi versus Replika. If the priority is ready photos/videos in chat and one-time billing, AISoul's photo workflow is a different option—but it has no published per-fact editor, no live voice/video calls and a finite media library. It is not evidence of better memory than Nomi.

For the duration question, read how long an AI companion remembers you. Do not rebuild a relationship elsewhere assuming that importing a character description also imports the old chats or learned state.

FAQ: Nomi AI questions before subscribing

What did the Nomi memory test actually find?

The cited Lab review labels its test July 24, ending July 19, and reports three evolving facts recalled at both day 7 and day 21 using web/mobile web and trial/paid use. It describes a calm text-first experience and warm voice notes. We did not reproduce that result or establish a fresh-account/Chrome/no-model-change condition. The denominator is three facts, not all future memories.

Does one 21-day test prove Nomi always remembers?

No. It reports a possible positive result in a particular dated test, not universal recall or a fully controlled cross-product benchmark. A wrong reply also cannot reveal the hidden cause on its own. Run your own neutral-cue return and correction checks with dates, plan and intervening message count recorded.

How much did Nomi cost in the captured test?

The July capture listed $15.99/month, $39.99/quarter and $99.99/year. The same prices appear in Nomi's March 20 official publication reopened October 5, but neither is a fresh personalized invoice. Confirm taxes, store/region differences, first charge and renewal in your account. The annual price is an upfront recurring yearly charge, not $8.33 billed monthly.

Is Nomi a visual-first AI girlfriend app?

The dated Lab experience was memory/text-led, but current documentation includes contextual selfies in chat, an album, art and image-to-video. Nomi is not “text only.” Whether those tools make it the right visual choice depends on the active image rules, allowance and results; this review has no matched photo-quality ranking.

Who is Nomi worth paying for?

Someone who values an ongoing relationship/roleplay, observes acceptable return recall and correction effort, and accepts the actual media and recurring-billing terms. The ledger turns those requirements into a keep-or-cancel decision. It does not require a universal session count, or mean Nomi cannot support multiple characters or media-focused use.

Can I erase an old fact by editing Nomi's Mind Map?

No. Nomi's correction guide says editing the summary does not change or delete underlying memories. Clarify the new fact in conversation, edit the exposed summary where appropriate, and check the later reply. Use account-deletion policies for data removal, not a memory-summary edit.

How much can I try for free?

Nomi's current wiki describes 50 messages/day on an indefinite free plan; its cited vendor comparison publishes two free images/day. Verify the displayed account allowance and desired tool before submitting. Free text access does not establish free voice or an unrestricted video allowance.

Are selfies the same thing as a live video call?

No. Nomi documents selfies and image-to-video generation as media workflows. Paid voice calls are another documented feature. A generated clip is not a live camera or two-way video call, and image credits alone do not establish a video price.

Does deleting the app cancel Nomi's subscription?

No. Cancel through the web, Apple or Google channel that billed you, complete the confirmation and record the end date. Cancellation, account deletion and editing a memory are separate actions. The official policies generally do not refund unused subscription time.

Dated testing, vendor claims and commercial disclosure

Evidence routes for reproducing this page's checks: Nomi versus Replika comparison; AI companion memory-duration test; AI companion landing page; AI Girlfriend Lab's dated Nomi review; Nomi's current product page.

All reported hands-on observations originate from the cited commercial Lab's dated test; the ledger and examples are proposed reader methods, not fabricated new results. Public primary documents reopened October 5 support current feature descriptions, published allowances and policies, not a guaranteed personal outcome or a newly captured invoice. AISoul has a commercial interest, and none of its quotas, Pass terms or media rules are used as Nomi evidence.

Primary sources are linked beside the claims: Mind Map announcement and correction wiki; Nomi 101, selfie and video guides; current voice/refund wiki; March 20 pricing/allowance publication; cancellation and April 27 privacy policies. Older publication dates remain older dates despite this review's October 5 update. No new comparative memory, image-quality, latency, medical-benefit, absolute-security or emotional-outcome result is claimed.