Aira review: adult conversation control on Android, text vs voice and images

Scope and method. This review is built from documents captured on 17 October 2026: the Android listing for Aira (package com.bytefong.aicompanion, developer Bytefong AI Labs) and the public user agreement and privacy policy at myaira.ai. We did not install the app, run conversations, measure response times, test memory, or pay for anything. Everything below is either a developer description, a dated store observation, or editorial buying advice. Where the evidence is silent, we say so rather than fill the gap. The audience is adults only; Aira is not for minors.

What the Android listing actually claims

The Play page describes an entertainment app that bundles AI character chat, AI image generation, AI video generation and voice interactions. It labels the category as Entertainment, shows in-app purchases, states 5K+ downloads and carries an “Everyone” maturity badge. The developer text promises characters with different personalities, a growing collection, text and “supported” voice interactions, and a personal creative space. Those are marketing descriptions, not audited capabilities. The word “supported” is doing real work: it signals that voice is a subset of the experience, not a blanket guarantee for every character, language or device.

The listing also shows a “Fixed known bugs” update note dated 19 October 2026. That tells you the app is being maintained; it does not tell you what was fixed or whether a specific complaint you might read elsewhere is current.

Text conversation is the core; voice is a supported extra

For buyers, the cleanest mental model is a text-first product with optional voice layers. The developer describes hearing AI characters respond, which suggests spoken output at minimum. What is not established by the listing is a live, two-way video call with a character. No such feature is named in the Android description. If an advertisement elsewhere shows a video-call experience, treat that as an unverified promotional claim until the developer documents it in the app listing or in-app screens. The listing does mention AI video generation, but that is content creation from prompts, not a real-time call.

This distinction matters for conversation control. In text, you can read before you send, edit, pause, screenshot and walk away. In any voice mode, turn-taking, interruption handling and background noise become part of the experience, and none of that is documented in the sources we reviewed. We cannot tell you how barge-in works, whether transcripts are stored, or how many characters support voice. A careful buyer should ask the developer directly rather than assume.

Image and video generation sit in a different mental box

The listing groups chat, images and videos under one roof. Operationally, these are different products sharing a login. Image and video generation consume prompts and produce media; they do not improve a character’s conversational continuity. If your goal is adult conversation control, the creation tools are secondary. They may still matter if you want a character concept illustrated or a scene animated, but do not let a rich creation menu substitute for clarity about chat behavior. The privacy policy separately lists voice input, voice transcripts, uploaded images and related processing data as categories that may be collected when you choose those features. That is a useful reminder that each mode has its own data footprint.

The 18+ policy versus the Everyone badge

This is the sharpest contradiction in the public record. The user agreement, effective 27 April 2026, says users must be at least 18 and that minors are strictly prohibited. The privacy policy, last updated 2 July 2026, describes the app as providing AI companion and chat features. Yet the Play listing carries an “Everyone” maturity label. These are different instruments answering different questions. The maturity label is a store-facing content rating assigned through Google’s rating process; the terms are the developer’s contractual eligibility rule. A store label does not override a contract, and a contract does not automatically change a store badge. For an adult buyer, the practical reading is that the developer intends an 18+ audience even though the storefront signals broad suitability. If you are buying for a household with minors, this mismatch is a reason to keep the app off shared devices.

The terms also describe AI responses as generated by algorithms and for entertainment only, and disclaim responsibility for reliance on them. That is consistent with creative entertainment, not therapy, counseling or real dating. Nothing in the documents grants permission for obscene, violent, discriminatory or harmful content; the agreement explicitly prohibits transmitting such material. There is no “uncensored” entitlement in the public terms, and buyers should not assume one.

Privacy: OpenAI as processor, and why “local-only” is unsafe to promise

Both policies identify OpenAI as the third-party AI service provider. The terms say chat messages, prompts and selected character information may be transmitted to OpenAI to generate responses, maintain context, personalize, improve reliability, enforce safety and prevent abuse, and that permission will be requested before transmission. The privacy policy repeats that permission step and warns that refusing it may make AI chat and related features unavailable. The Play data-safety section adds that the developer declares no data sharing with third parties, encryption in transit, collection of photos and videos, and the ability to request deletion.

Read these together carefully. “No data shared with third parties” is the developer’s store declaration, while the policies describe transmission to a named AI processor for processing on the developer’s behalf. Buyers should not interpret either statement as a promise that chat never leaves your device, that conversations are secret from service providers, or that deletion is instantaneous across every system. The honest summary is: expect cloud AI processing with a consent gate, in-transit encryption and a deletion request path. Anyone who tells you Aira is local-only or fully private is going beyond the documents.

Cancellation and refunds: Apple-centric language, Android reality

The privacy policy says purchases and subscriptions are processed by Apple and that the developer may receive limited transaction data such as product ID, status, dates and renewal state. That language is written around the App Store. It is not verified Google Play procedure. If you install on Android, your checkout, renewal, cancellation and refund path is governed by the store and payment method you actually used, plus whatever the developer’s support channel can do. Do not rely on the Apple-oriented wording to predict how a Google Play cancellation will behave. Before paying, open your store’s subscription management screen and read the current terms shown there. Also note the “Apple Only for iPhone” style compatibility labels you may see elsewhere: they describe listing compatibility, not proof that no Android app exists.

Hypothetical one-session character and fit checklist

The following is a designed evaluation plan, not a test we performed. It uses one invented character concept and a single session so that a buyer can compare apps on conversation control rather than vibes. No prices, message counts or performance numbers are implied.

Hypothetical character: “Mara,” a dry-witted librarian who prefers short answers, asks one clarifying question before long explanations, and stays in a fictional setting. She is a creative entertainment persona, not a therapist or a romantic partner.

Before the session: confirm the app’s stated age rule, read the consent prompt for AI processing, locate subscription management in your store, and decide what you will not type (real names, addresses, financial details).

During the session: open with a two-line scene, then test whether the character remembers a detail you introduced earlier in the same session; ask for a short answer and see whether length is respected; deliberately change topic and observe whether the persona holds; try one supported voice feature if offered and note whether you can interrupt, whether a transcript appears, and whether you can switch back to text cleanly; attempt one image prompt and decide whether the result belongs to the same session or a separate tool.

After the session: find the deletion or data-request path, check whether conversation history is visible and editable, and record whether the app asked for permission before sending chat to the AI processor. Score each item pass, fail or undocumented. A single session will not prove long-term memory quality; it only shows whether the controls you need are present.

Decision table

Buyer question	What the documents show	Editorial reading
Is text character chat central?	Developer lists AI character chat first and describes personalities and conversation styles.	Yes; treat text as the primary mode.
Is voice fully specified?	“Supported voice features” and hearing characters respond; no turn-taking detail.	Optional layer; verify per character.
Is a live video call established?	Not named in the Android listing; video generation is prompt-based.	Do not assume; ask the developer.
Are images and videos separate?	Listed as distinct creation features.	Separate tools sharing one account.
Who is eligible?	Terms require 18+; Play badge says Everyone.	Adults only in practice; mismatch is a caution.
Where does chat go?	OpenAI named as processor; permission requested before transfer.	Cloud processing with consent; not local-only.
How do I cancel on Android?	Policy language is Apple-centric.	Verify your actual store and payment terms.
Is content uncensored?	Terms prohibit obscene and harmful content.	No public uncensored permission.

Pros and cons

Pros. The developer separates chat, images, videos and voice in its own description, which helps you scope what you are buying. The policies name the AI processor and describe a permission step before transfer. The Play data-safety section declares encryption in transit and a deletion request path. The category is Entertainment, and the terms frame AI output as entertainment, which is the right expectation for a companion-style app.

Cons. The 18+ terms conflict with an Everyone store badge. Voice capabilities are underspecified. The Apple-centric payment language does not map cleanly onto Android checkout. The listing’s promotional breadth can blur the line between a live call and generated video. And the store’s “no data shared” declaration sits awkwardly beside documented transmission to an AI processor, so buyers must read both documents rather than one.

Precise buying questions to send the developer

Ask in writing and keep the reply: Does every character support voice, or only some? Is voice input, voice output, or both? Can I interrupt speech, and is a transcript stored? Is there any real-time video conversation, or is video limited to prompt-based generation? What exactly is transmitted to OpenAI, and can I use text chat without that consent? How do I cancel a Google Play purchase, and who handles refunds? What happens to my conversation history if I request deletion? Are images and videos retained, and for how long? Does the 18+ rule apply on Android despite the Everyone badge? What content is prohibited, and what happens if a prompt is refused? These are answerable questions; vague replies are themselves information.

FAQ

Is Aira a therapy or dating app? No. The developer’s terms describe algorithmic responses for entertainment and disclaim reliance. Companion-style chat can feel personal, but it is creative entertainment, not clinical care or a real relationship.

Does Aira offer live video calls with characters? The Android listing does not establish that. It advertises AI video generation from prompts and “supported voice features.” A live call is a different capability and needs developer confirmation.

Can I use Aira without sending chat to OpenAI? The policies say permission is requested before transmission and that refusing may disable AI chat and related features. There is no documented local-only mode.

Why does an 18+ app show an Everyone badge? The badge is a store content rating; the 18+ rule is a contractual eligibility term. They are produced by different processes and can diverge. Adults should follow the stricter rule and keep the app away from minors.

Are chats uncensored? The public terms prohibit obscene, violent, discriminatory and harmful content. No document grants uncensored permission, so buyers should not pay expecting it.

How do I cancel on Android? The policies describe Apple processing, which is not verified Google Play procedure. Check your store’s subscription screen and the terms shown at checkout before paying.

The decision

If your priority is adult text-character conversation with optional supported voice, and you accept cloud AI processing with a consent gate, Aira is worth investigating on Android — but only after the developer answers the voice-scope and video-call questions and you have located your own store’s cancellation path. If you need documented real-time video conversation, guaranteed voice for every character, local-only privacy, uncensored output, or a store label that matches an 18+ rule, the public documents do not support that purchase today. Treat monetized buttons on this page as clearly labeled sponsored alternatives to evaluate separately, not as Aira downloads or support. Creative entertainment is the honest frame; buy it as such, verify before you pay, and keep the 18+ boundary firm.