Talkie is a character chat platform where you converse by text or voice with AI personas built mostly by other users, powered by MiniMax models and operated by Singapore-based SUBSUP, with strict published content rules, a card-trading economy and creator revenue sharing.
Used this tool? Rate it
Used this tool? Rate it
Talkie is a character chat platform where you converse with AI personas — some built by the company, most built by other users — through text and voice. The official site titles itself "Talkie | Free AI Character Chat" and invites you to "chat with AI characters for free" and "discover your perfect AI friends."
Two facts about this product are worth establishing before anything else, because both are commonly reported incorrectly.
The first concerns who makes it. The official FAQ states that the characters "are purely the product of AI technology power by Minimax and are not controlled or intervened by real humans." MiniMax is the Shanghai AI company whose models drive the experience. However, the entity you actually contract with is different: the privacy policy names SUBSUP PTE. LTD. as the data controller, the Google Play listing names SUBSUP as the developer with an address at 8 Kaki Bukit Avenue 4 in Singapore, and support runs through feedback@subsup.com. Model provider and operating company are not the same, and it matters for questions about data and jurisdiction.
The second concerns what kind of product this is. Talkie is widely filed under "AI companion," a category that carries an assumption of adult content. The official Community Guidelines say the opposite in unusually direct language. Under the heading Adult Content, they state that the platform prohibits "both explicit sexual content and content with suggestive or implicit sexual undertones," and then enumerate what that covers: nudity or sexually suggestive imagery, pornographic or sexually explicit materials in any form, sexual exploitation or non-consensual scenarios, sexually suggestive poses or descriptions, and content that objectifies or inappropriately sexualizes individuals. The FAQ repeats the point in its own NSFW section, stating that pornographic and violent content are prohibited per the terms of service while the platform aims "to support a diverse range of character types within safe parameters."
This is not a case of a policy on paper being ignored in practice. The most striking pattern in user reviews, covered in detail below, is that people complain the filters are too tight rather than too loose.
There is one more identity note. It is easy to conflate this product with similarly named apps. The US App Store returns several unrelated products for the query, including TalkZ from Going Merry LLC and Talk.AI from BE.AI. Everything on this page is anchored to three identifiers together: the domain www.talkie-ai.com, the publisher SUBSUP PTE. LTD., and the Android package com.weaver.app.prod.
A note on the listing data that brought you here: the site's stored title and description in every language were the text of a Cloudflare interstitial — "Just a moment..." and its translations. That was a capture artifact rather than real product copy, and this page is written from the live site instead.
The core loop is conversation with AI personas. The Google Play description positions the product as a creation community rather than a companion app, describing it as a place where users "craft, customize, and collaborate with dynamic AI personas" and explore other people's user-generated content. The app's own name reflects that framing: it is now listed as "Talkie: Creative AI Community," having previously carried the "Soulful AI" label — a shift in wording from companionship toward creation.
Characters are described as evolving with extended memory as you engage with them, adapting to become more personalized over time. The company is candid about the ceiling here. Its FAQ states plainly that "AI memory capacity is technically limited, causing older conversations to be 'forgotten,'" and says work is underway to retain more history. Independent hands-on testing found characters reliably held planted details across more than twenty messages in text mode, but that details established in text sometimes did not survive a switch into voice mode.
The homepage surfaces a category of characters that are straightforwardly functional rather than social: a French teacher, a mock interviewer, a personal nutritionist, an industry insider for career questions, an ADHD personal helper, and a trend analyst. There is also a simulation genre — running a Japanese izakaya, or navigating a superpower academy — where the appeal is scenario management rather than relationship. Anyone assuming this product is only romantic roleplay is working from an incomplete picture.
The FAQ describes a layer that distinguishes Talkie from most competitors: users earn card rewards based on the topics they discuss, and those cards can be traded between users for virtual currency, which buys further rewards. The company also runs revenue sharing, stating that creators using its user-generated content tools "can earn revenue shares," subject to originality and copyright compliance, with income depending on user transactions and payments. Notably, Apple's own advisory for the iOS app flags Loot Boxes — a direct consequence of this mechanic.
Unusually, moderation is documented as a user-facing behavior rather than hidden infrastructure. The FAQ explains that all AI settings undergo review before taking effect, which "can sometimes delay updates or result in temporary removal if flagged as inappropriate," and that characters violating community standards may be temporarily hidden pending review. The company says it combines human oversight with AI technology and provides an appeal path through its feedback address. This design choice explains a large share of the negative reviews, and understanding it upfront will save you frustration.
The largest use is open-ended story play with characters others have written — historical settings, fantasy scenarios, slice-of-life premises. Individual characters display their interaction counts, so popular ones are easy to find.
Because creation tools are paired with revenue sharing and a public library, some users treat the platform as a publishing venue rather than a private toy. That reframes moderation from an annoyance into a business risk, since a hidden character means a lost audience.
The French teacher and mock interviewer personas point at a genuine strength of character chat: practicing a conversation where mistakes cost nothing. Given the company's own caution about reasoning limitations, this works better for fluency drilling than for accuracy-critical instruction.
The plain social use — talking through a day, having something respond warmly — remains a real draw, and the memory system makes continuity feel meaningful within a session.
It is not suited to unfiltered adult roleplay, as the guidelines and the review evidence both make clear. It is not suited to work requiring factual reliability, since the company itself warns responses may be inaccurate or unverified. And it is not suited to anyone who needs their creative work to stay published without review.
Treat the content policy as a design constraint, not an obstacle. The guidelines prohibit suggestive undertones and not merely explicit content, so scenarios built around romantic tension will run into redirection. Users who plan around this are far happier than users who fight it.
Front-load the details you want remembered. Memory is limited by the company's own admission and degrades with distance. Establishing key facts early and restating them at intervals produces better continuity than assuming persistence.
Expect a review delay when creating. Since all settings pass moderation before taking effect, an edit that seems not to have applied has most likely not finished review. Waiting is more productive than repeatedly resubmitting.
Do not rely on it for facts. The FAQ explicitly warns that responses can be inaccurate or unverified and advises seeking professional advice for serious matters. Treat outputs as fiction by default.
Understand the refund position before spending. The FAQ states recharges are generally non-refundable except for system errors or duplicate transactions. Buy the smallest useful amount first.
If you are building an audience, keep your work backed up. Reviewers report characters and profile images being removed without a clear explanation. A creator treating this as a publishing platform should not keep the only copy inside it.
Talkie fits people who want conversational roleplay and character creation inside a moderated environment, who are comfortable on Android or the web, and who value a large user-generated library over deep customization freedom. Creators interested in revenue sharing and an audience are a distinct and well-served group.
It fits poorly for anyone seeking unfiltered adult interaction, anyone who needs factual reliability, anyone unwilling to tolerate advertising on the free tier, and — given the age threshold of 14 and the presence of user-generated content — this is a product where parents should look at the actual content policy and the divergent store ratings rather than assuming either the best or the worst.
Talkie runs on the web and on Android. The Google Play listing shows the app as "Talkie: Creative AI Community" from developer SUBSUP, updated on 23 August 2026, with more than 50 million downloads, containing ads and in-app purchases.
The iOS situation requires explanation and is the single most important availability fact about this product. In December 2024, the South China Morning Post reported that Talkie had disappeared from the US App Store, with the team citing unspecified "technical reasons" and saying "We will get it fixed soon for you" without a timeline. At the time of this writing, a search of the US App Store still does not return Talkie itself. What it does return is "Talkie Lab - AI Playground" from the same publisher, SUBSUP PTE. LTD., rated 4.73 from 55,285 ratings, at version 2.58.001 released on 20 August 2026, requiring iOS 15.1 or later.
So iOS users in the US are not using the same listing that Android users are. That distinction matters for ratings comparisons and for anyone trying to reconcile conflicting reports about whether the app is available.
Honest disclosure on this section: the vendor does not expose a directly retrievable pricing page. Several conventional paths returned nothing usable, so no subscription figures are asserted here. What can be verified first-hand is the following.
The product is free to start. The site's navigation carries an "Unlock with Talkie+" entry alongside a 50 percent discount marker, indicating a paid tier exists under that name. The Google Play listing confirms the app contains both advertising and in-app purchases. Free-tier advertising is heavy enough to be a recurring review complaint, with one user describing three ads every ten minutes, and another noting that in the current version text already typed is lost when an ad interrupts, whereas an older version preserved it.
On refunds, the FAQ is explicit: recharges are generally non-refundable, with exceptions only for system errors or duplicated transactions.
There is also money flowing the other direction. Creators using the platform's user-generated content tools can earn revenue shares, provided content is original and copyright compliant, with income depending on user transactions and payments. Combined with the card-trading economy, this makes the product's internal economy more elaborate than a simple subscription.
For current prices, check the app or website directly rather than relying on any figure quoted second-hand.
Character.AI is the closest and best-known comparison, similarly built on user-generated characters with its own moderation regime. PolyBuzz and Chai occupy adjacent territory on mobile with large rating counts of their own. Janitor AI and similar platforms serve users specifically seeking looser content policies — which is precisely the gap that Talkie's own reviewers point to when they say competitors have "gotten far looser on their censorship." Replika targets sustained one-to-one companionship rather than a browsable character library.
The selection question is unusually clean here. If your priority is a large library inside a moderated environment with a creator economy attached, Talkie is a strong fit. If your priority is content freedom, the guidelines quoted above tell you plainly that this is the wrong product, and no amount of prompt technique will change a policy that is enforced at review time.
Content filtering is strict and is the dominant source of user dissatisfaction. This deserves stating plainly because it inverts the usual assumption about the category. Reviewers report being told to "please change your topic" repeatedly, being unable to create characters because unnamed words trigger policy without the system explaining which, and having profile images removed without clear justification. One reviewer summarized the competitive position bluntly, saying rivals "have for the most part gotten far looser on their censorship, and Talkies stagnation is leaving it in the dust." Independent testing describes filters that work by gently redirecting a conversation back to safer ground rather than refusing outright, permitting light flirtation while blocking explicit content.
Moderation transparency is a recurring complaint. Several reviewers describe content flagged and removed without a clear reason. One account describes losing an audience of roughly 800 followers after a community page takedown and account deletion. Sample sizes are small, but the theme is consistent and follows directly from the documented design in which everything passes review.
Memory is limited by the vendor's own account. Older conversations are forgotten, and independent testing found that details established in text mode sometimes did not carry into voice mode.
Reasoning is limited by the vendor's own account. The FAQ warns of limitations on topics requiring complex logical reasoning and states that responses can be inaccurate or unverified.
Advertising on the free tier is heavy. Reviewers describe frequent interruptions, loss of typed text when an ad appears, and no offline use.
Store ratings diverge sharply and must be read together. Google Play shows 4.6 from 832,000 reviews across more than 50 million downloads — by far the largest and most meaningful sample. The iOS Talkie Lab listing shows 4.73 from 55,285 ratings. Trustpilot shows 2.9 from 56 reviews on an unclaimed profile, distributed at 23 percent five-star, 13 percent four-star, 20 percent three-star, 14 percent two-star and 30 percent one-star. The Trustpilot sample clears the twenty-review threshold but is four orders of magnitude smaller than the Play Store's, and the gap between a 4.6 and a 2.9 is best read as the difference between a broad user base and a self-selected complaint channel rather than as a contradiction to be resolved.
Age and content ratings differ across stores, and both should inform your judgment. Google Play rates the app Teen, with content descriptors that explicitly list Sexual Themes alongside Users Interact and In-App Purchases. Apple rates the iOS Talkie Lab listing 17+, and its only advisory is Loot Boxes rather than sexual content. The privacy policy sets its own floor at 14 years of age, stating the services are not intended for individuals under 14 or under the applicable local minimum, and says the company verifies the age of registered users without describing the method — so no conclusion should be drawn about whether meaningful age verification exists. Three different thresholds from three sources, pointing at the same product.
A third-party category label points the other way. Trustpilot files the domain under "Adult Entertainment Club" and "Adult Dating Service." That is an automated third-party classification and it contradicts the company's own published guidelines. It is reported here rather than reconciled, because both facts are true simultaneously: the platform's policy forbids adult content, and at least one external directory has categorized it as adult.
The operating entity is not the model provider. MiniMax supplies the technology; SUBSUP PTE. LTD. in Singapore is the data controller and the party in the store listings.
Availability on iOS in the US is unresolved. The original app left the US App Store in December 2024 for reasons the company described only as technical, and it has not returned under that name.
No, and the policy is more restrictive than most people expect. The Community Guidelines prohibit both explicit sexual content and content with suggestive or implicit sexual undertones, listing nudity, pornographic material in any form, sexually suggestive poses or descriptions, and content that objectifies individuals. The FAQ states separately that pornographic and violent content are prohibited per the terms of service. The strongest confirmation is not the policy text but the review pattern: users complain overwhelmingly that filtering is too aggressive, not that it is too permissive.
Two entities are involved. The official FAQ states the characters are powered by AI technology from MiniMax, the Shanghai company behind the models. The operating company is SUBSUP PTE. LTD., named as data controller in the privacy policy, listed as the developer on Google Play with a Singapore address, and reachable at feedback@subsup.com. Contracts, data handling and support run through SUBSUP.
The privacy policy states the services are not intended for individuals under 14, or under the applicable local minimum where that is higher, and says accounts below the threshold will be restricted and deleted once identified. The same policy lists verifying the age of registered users among its purposes but does not describe how, so it should not be assumed that a substantive verification step exists. Note that the store ratings use different thresholds: Google Play rates the app Teen while Apple rates the iOS Talkie Lab listing 17+.
Not straightforwardly in the United States. The South China Morning Post reported in December 2024 that the app had vanished from the US App Store, with the company citing unspecified technical reasons and offering no timeline. A US App Store search still does not return Talkie itself; it returns "Talkie Lab - AI Playground" from the same publisher, rated 4.73 from 55,285 ratings. Android availability through Google Play is unaffected.
Yes to start, with costs attached. The Google Play listing confirms the app contains ads and in-app purchases, and the site's navigation shows an "Unlock with Talkie+" paid tier. Free-tier advertising is frequent enough to be a common complaint. No subscription prices are quoted on this page because the vendor does not expose a directly retrievable pricing page — check the app or site for current figures.
Because they are being reviewed. The FAQ states that all AI settings undergo moderation to ensure they meet community standards, which can delay updates or result in temporary removal if something is flagged. Characters violating standards may be temporarily hidden pending review, and you can revise the content for re-evaluation or contact support. This is documented product behavior rather than a malfunction.
Limited, by the company's own account. The FAQ states that memory capacity is technically constrained and that older conversations are consequently forgotten, with improvements said to be in progress. Independent hands-on testing found characters held planted details reliably across more than twenty messages in text mode, but that details established in text sometimes failed to carry over when switching to voice mode.
Read them together. Google Play shows 4.6 from 832,000 reviews with more than 50 million downloads, which is the only sample large enough to carry real statistical weight. The iOS Talkie Lab listing shows 4.73 from 55,285 ratings. Trustpilot shows 2.9 from just 56 reviews on an unclaimed profile with a notably scattered star distribution. The small Trustpilot sample functions as a complaint channel and should be read for its themes rather than its score.
The FAQ states that creators using the platform's user-generated content tools can earn revenue shares, with content required to be original and copyright compliant, and income depending on user transactions and payments. There is also a card economy in which conversation earns card rewards that users can trade for virtual currency. If you pursue this seriously, weigh it against the moderation risk described above, since removal of a character removes its audience.
Generally no. The FAQ states that recharges are non-refundable except in cases of system errors or duplicated transactions, and directs users to review the terms and contact customer service. Given that policy and the advertising-heavy free tier, trying the free experience thoroughly before spending is the sensible sequence.
