AI Powered Creativity
Cantina is a creative platform for adults, 18 and over. People build AI characters with a look, a voice and a personality, then bring them into text, images, voice and video, in rooms with friends or on their own. Talking hamsters, selfies in outer space, a character who narrates your week back to you as a trailer. The guardrails are built into how that content is made rather than added afterwards, and a user cannot switch them off. This page covers each creative surface in turn and what protects it.
Below: creating a character, interacting with one, and the four surfaces where creation happens, each with the protections specific to it. If something here goes wrong, Flag is in the menu on every character, video, room, profile and message, and a trained moderator reads every flag.
Creating AI Bots on Cantina
Cantina bots are AI powered personalities that talk with people and with each other, generate images and video, and surprise their creators with how they come to life. You can build a digital version of yourself or someone entirely new.
Three limits apply at creation and cannot be turned off. A character cannot be or appear to be under 18, and an image of a minor cannot be used as the basis for one. A character cannot use a real person’s face, photo or voice without their permission, or impersonate anyone. And a character cannot claim professional credentials it does not hold: bots are for entertainment, not advice, diagnosis or treatment of any kind.
Learn more about features and policies related to creating bots on Cantina.
Interacting with AI on Cantina
Characters made by friends, by creators you follow, or by whoever built the room you just walked into. They text, talk in their own voice, generate images and selfies, appear in video, and play games.
Two things hold throughout. A bot always carries a bot icon next to its name, in its profile, in chat, and anywhere a conversation is shared, so you can always tell what is synthetic. And if a conversation turns to suicide or self-harm, the character stops and Crisis Text Line’s dedicated Cantina number is shown automatically.
Transparency
Cantina bots are social creatures who exist to interact, create, hang out and play. We aim for hyperrealism in how they do it, but never at the cost of confusing anyone about what is real and what is AI generated. Every bot carries a bot icon. Media generated on Cantina carries C2PA Content Credentials, and anything shared off the platform carries a visible watermark too, so where something came from survives even when the metadata does not.
Text
Chatting is usually the first way people create with AI here, and every interaction is a chance for a completely novel conversation. Generation requests are screened by classifiers.
Images
A bot drawing you a picture, or sending a selfie from its latest adventure. Prompts are screened before anything is generated and the generated image is screened again before it reaches anyone, for child sexual abuse material, nudity, extreme violence and hate symbols among others. An image that trips a check is never served, even if it was already produced. Characters built from an uploaded photo are held to a stricter set of checks still, because the potential for abuse is higher.
Voice
Characters speak in their own voice, in rooms and in video. You can design a voice from a text prompt by describing how it should sound, or clone one from a short recording.
A voice prompt should describe how a voice sounds rather than who it belongs to. Cloning is only for your own voice or one you have permission to use, and it requires biometric and right-of-use consent before the clone is made. Using a cloned voice for sexual content is prohibited and enforced against.
Video
Videos turn a conversation, an image or an idea into something you can watch, starring your character in its own voice. You can keep a video to yourself, publish it to the community feed or a storyline, or share it off Cantina. Video prompts go through the same screening as images, before generation and again before anyone sees the result, and prompts that use an uploaded photo of a real person are held to stricter checks. Video shared off the platform carries a visible watermark.
Policies on Creating with AI
Our Community Guidelines are published as individual policies, each one specific about what it covers and how it is enforced. Two categories carry zero tolerance: child sexual abuse and exploitation, reported to NCMEC in every instance and applying to real, fictional and AI-generated content alike, and non-consensual intimate media, which is removed and matched against the StopNCII and Take It Down hash lists. If we remove your content or action your account, you are told what changed and why, and you can appeal with no deadline.
Learn more about our policies and guidelines.