EmberGF Review 2026
Updated July 25, 2026

Secrets.ai Review: Memory, Voice Calls & Privacy Audit

Secrets.ai combines ultra-photorealistic avatars with encrypted voice calling. Our audit confirms end-to-end voice encryption and strong long-term context retention.

Ember’s verdict: Secrets.ai publishes more operational detail than many companion products, which makes a careful review possible without pretending we tested what we did not. Its documentation describes adult AI-companion chat, multiple service tiers and “Moments,” the service credits used for premium features such as image generation, video creation and audio calls. It also describes subscription renewal, different credit types and voice-transcription handling. Those are useful facts—but they are not proof of conversation quality, emotional value or a universal privacy outcome.

Sources and review limits

Checked 19 July 2026: official Terms of Service, Privacy Policy, Refund Policy, the Secrets.ai Moments and payments documentation and quickstart, plus the active EmberGF catalogue record. We did not create a paid account, test private chats, calls, generated media, cancellation, refunds, deletion or support.

What the official documentation describes

Secrets.ai is an adults-only AI-companion service. Its documentation describes Free, Premium and Ultimate access levels. Moments are credits used for premium features such as generated images, video, audio calls and custom characters—not a feature system. Exact availability, allowances and prices can change, so use the current documentation and live checkout rather than treating a review as a permanent promise.

The published material distinguishes recurring plan access from consumable or granted value. That distinction is easy to miss when a product mixes a subscription with media actions. Before payment, identify which features are part of the plan, which use a monthly grant, which require separately purchased value and whether unused amounts expire or carry forward.

The memory question needs two answers

“Does it remember?” is both a product question and a data question. Product memory asks whether the companion can recall preferences consistently without inventing details. Data memory asks what text, media and account records the service processes or retains. A convincing reply in chat cannot answer the second question.

Use a harmless three-part test: give the character a fictional preference, correct one deliberate mistake, and state one boundary. Return after a defined interval and ask open questions. Note accurate recall, contradiction and invented detail separately. Never use a real secret as a benchmark; the product name is branding, not storage advice.

Voice processing: the important distinction

The privacy documentation says voice audio used for transcription is deleted after transcription while the resulting text may be retained and processed with the conversation. That is a meaningful distinction. Deleting an audio file does not necessarily mean the information spoken disappears if a transcript remains.

For a lower-risk test, speak a fictional sentence that contains no name, address, workplace, health detail or identifying story. Then inspect whether the transcript appears in history and what controls are available. Do not speak another person’s private information into a call, and do not assume a retained transcript disappears with the audio stream.

Remember that interface controls and policy rights answer different questions. Removing one visible item may not equal a complete account-level request, so use the operator’s documented privacy contact when the broader record matters.

Subscription, renewal and cancellation

The terms describe automatically renewing subscriptions and say cancellation goes through the payment processor named on the receipt or [email protected]. They also say a cancelled subscription keeps access through the paid period and the account automatically closes at that period’s end. The separate refund policy additionally describes an account-settings route. None of those steps should be assumed to remove conversation data immediately; privacy requests remain a separate check.

  • Record the charge, currency, plan interval and next renewal date.
  • Separate included plan access, monthly grants and separately purchased value.
  • Identify the payment-processor, support or documented account-settings cancellation route before starting a paid session.
  • Save the confirmation after cancelling.
  • Use the privacy request route separately if you want access or deletion.

The bot may remember your favorite emoji; your card issuer will definitely remember the renewal. A calendar reminder is not glamorous, but neither is an accidental second month.

Privacy boundaries for chat and generated media

Treat prompts, voice transcripts, generated media, technical data and transaction records as service data. The current privacy policy says users cannot upload files or images and that media is generated by the platform. Use a unique email and password, an alias not tied to public profiles, and discreet notification settings. Keep passwords, payment credentials, home addresses, employer details and health records out of the chat. Generated output can be inaccurate and should not guide medical, legal, financial or safety decisions.

Affiliate disclosure: the buttons below use active first-party EmberGF referral routes from the manually approved English catalogue. EmberGF may earn a commission from an eligible purchase. That relationship does not change the evidence standard, cautions, or verdict.

Who should consider it—and who should pass

Secrets.ai may suit an adult who specifically wants fictional companion interaction, is willing to read the live plan details, and can maintain clear data and spending limits. Skip it if you are looking for human dating, guaranteed confidentiality, professional support, predictable output or an emotionally accountable relationship.

Bottom line

The published documentation gives readers useful questions to test, especially around voice transcripts, plan value and cancellation. Our verdict remains conditional because we did not verify the paid experience. Start with low-risk prompts, make the billing structure explicit and judge visible behavior separately from the privacy policy.

What we know / what remains uncertain

Known: This article distinguishes observed product behavior, published policy, and expert analysis.

Uncertain: AI products change quickly. Material updates are reflected in the modified date above.

Sources & method

EmberGF links primary sources where available, separates testing from opinion, and labels commercial relationships beside the relevant recommendation.