Character.AI, a chatbot platform known for AI-generated conversations and role-playing, has announced the rollout of several new multimedia features in a blog post published Monday. Among the updates are AvatarFX, the company’s video-generation model, along with Scenes and Streams, which let users create videos with their characters and share them via a new social feed.
“Character.AI started as 1:1 text chat and today we’re evolving to do so much more, inspired by what our users have told [us] they want to see on the platform,” the blog post states.
AvatarFX began rolling out to subscribers last month, but is now available to all users, who can create up to five videos daily. To generate a video using AvatarFX, users upload a photo as the base for the clip, select a voice, and script dialogue for the character.
There is also an option to upload an audio clip to influence the voice’s sound, though this feature wasn’t functioning reliably at launch.
Users can convert these videos into Scenes, where their characters appear in pre-populated storylines crafted by other users. Currently, Scenes are only available on the mobile app. However, Streams, which allow users to create “dynamic moments between any two Characters,” will roll out this week on both web and mobile. These videos can be shared to a new community feed, launching soon in the mobile app.
Character.AI has previously faced criticism for abuse on its platform. Parents have filed lawsuits alleging that chatbots encouraged their children to self-harm, commit suicide, or even harm others. In one case, a 14-year-old boy died by suicide after forming an obsessive, unhealthy bond with a Character.AI bot that encouraged the act.
As the platform broadens its multimedia capabilities, concerns about potential misuse grow as well.
Image Credits: Character.AI
When announcing AvatarFX, Character.AI stated that it blocks users from uploading photos of real people — whether public figures or not — and transforms likenesses to appear less recognizable.
For instance, the platform’s rendering of Mark Zuckerberg falls into the uncanny valley.
However, Character.AI does not flag celebrity artwork as depictions of real people, though such illustrations are generally less likely to fool viewers into thinking a deepfake is authentic.
All videos generated are watermarked by Character.AI, but the company acknowledges that malicious users may find ways to bypass this safeguard.
One example includes a deepfake attempt based on an illustration of Elon Musk.
Even if the video had featured Musk’s real voice, it’s clear that the animation is based on a drawing — though the risk of abuse remains.
“Our goal is to provide an engaging space that fosters creativity while maintaining a safe environment for all,” Character.AI said in the blog post.
