Overview
Candy AI is an AI companion platform centered on a single ongoing relationship with a character rather than browsing a large public catalog. A user picks one of the platform's pre-built characters or builds one, then returns to the same chat thread over time; voice messages, on-request images, and short video clips are layered on top of that core text conversation rather than offered as separate standalone tools.
Customization
Character creation covers physical traits (ethnicity, age range, eye color, hairstyle, hair color, and body features) plus personality traits such as bubbly, calm, cheeky, nurturing, or intellectual, along with voice preference and stated hobbies or interests. This upfront customization is closer to building a defined persona than picking from a fixed character roster, and it happens once, through a builder, rather than through ongoing in-chat editing.
Memory and context
Candy AI states that its companions "learn through conversation" and adjust their responses and tone to better match the user's own communication style over time, making later chats feel more personalized. The platform does not publish technical detail about how much conversation history is retained or how memory is structured across sessions.
Interface and generation workflow
Image and video requests are made inline with the conversation rather than through a separate creation studio: a user asks the companion for a photo or short clip, and the platform generates one reflecting the current scene. Candy AI describes its video output as reacting to user input in real time, which differs from platforms where video is a static, after-the-fact export. Voice works the same way: a user can request a spoken reply instead of text, and the companion answers in one of several voice styles, described by the platform as ranging from "soft and comforting" to "bold and confident."
The token economy
Text chat is covered by the subscription itself, but image, voice, and video generation draw from a separate token balance that refills each billing cycle and can be extended with additional token purchases. This two-layer structure (subscription for access, tokens for media) means the practical cost of using Candy AI depends heavily on how much image and video generation a user does, not just which subscription tier they hold.
Content policy and moderation
NSFW interaction is part of the default experience for verified adult users rather than a separate toggle. Candy AI's published Community Guidelines require users to be at least 18 (or the age of majority in their jurisdiction), and its Blocked Content Policy states a zero-tolerance stance on content involving or resembling minors and on non-consensual intimate imagery of real people, alongside standard prohibitions on hate speech, illegal content, and glorification of violence or self-harm. The platform describes screening conversations and generation requests with automated detection, with flagged content routed for review and enforcement ranging from content blocks to account suspension; content identified as child sexual abuse material is described as being reported to the relevant authorities rather than only removed.
Candy AI's guidelines describe age-gating as based on user attestation at signup rather than a documented identity- or document-verification step.
Privacy and technical claims
Candy AI's own materials describe the service as using end-to-end encryption and GDPR-compliant data-handling practices. Beyond that, the platform does not publish technical detail on model architecture, image resolution, or video length and frame rate, so those specifics aren't independently verifiable from official sources.
Notable features
Voice interaction is treated as a first-class part of the product rather than an add-on, and video generation (short AI-generated clips of the companion that respond to user input) is a further point of differentiation, since most character chat platforms in this space are limited to text and static images.
Limitations
No public API is documented for Candy AI, and there's no published technical specification for image resolution, video duration, or generation speed.
Important distinctions
Where Nomi bundles chat, voice, and image/video generation into one flat subscription tier, and Kindroid uses stackable subscription add-ons to scale memory and audio capacity, Candy AI keeps media generation on a separate, consumable token balance layered on top of the base subscription. Candy AI's emphasis on one deep, multimedia-rich companion relationship also sets it apart from community-character platforms like SpicyChat or CrushOn.AI, whose characters are largely built and published by their own user communities rather than created within the platform's own single-character builder.