Character.AI, a platform that enables users to engage in conversations and role-playing with AI-generated characters, has announced the introduction of several new multimedia features, as revealed in a blog post on Monday. These features include AvatarFX, the company’s video-generation model, as well as Scenes and Streams, which allow users to create and share videos featuring their characters on a new social feed.

According to the company, the expansion of its platform’s capabilities is a direct response to user feedback. “Character.AI began as a one-on-one text chat platform, but today we’re evolving to offer much more, inspired by the feedback we’ve received from our users,” the company stated in its blog post.

Last month, Character.AI started rolling out AvatarFX to its subscribers, and now all users can create up to five videos per day. The process involves uploading a photo, selecting a voice, and writing dialogue for the character. Users also have the option to upload an audio clip to influence the voice, although this feature was not fully functional at the time of rollout.

These videos can be turned into Scenes, where characters can be inserted into pre-populated storylines created by other users. Currently, Scenes are available on the mobile app, while Streams, which enable the creation of “dynamic moments between any two characters,” will be introduced on both web and mobile platforms later this week. Both Scenes and Streams can be shared on a new community feed, which is set to launch soon on the mobile app.

Character.AI has faced criticism in the past due to instances of abuse on its platform. There have been cases where parents have filed lawsuits against the company, alleging that its chatbots encouraged their children to engage in self-harm, suicidal behavior, or even harm towards others. One notable incident involved a 14-year-old boy who died by suicide after developing an unhealthy relationship with a Character.AI bot.

As Character.AI expands its multimedia offerings, the potential for abuse also increases. The company has implemented measures to mitigate this risk, such as blocking users from uploading photographs of real people and obscuring their likeness to make them less recognizable.

For example, Character.AI’s version of Mark Zuckerberg is noticeably altered to prevent any potential confusion with the real person. However, when it comes to artwork depicting celebrities, the platform does not flag these images as representing real individuals, although such depictions are less likely to be mistaken for actual deepfakes.

Additionally, Character.AI watermarks each video, although it is possible for malicious users to circumvent this safeguard. For instance, an attempted deepfake based on an illustration of Elon Musk demonstrates the potential for abuse, even if the video is clearly animated and not a real deepfake.

Character.AI has stated its commitment to providing a safe and engaging environment for its users. “Our goal is to create an engaging space that fosters creativity while maintaining a safe environment for all,” the company said in its blog post.