Character.AI Now Generates Video & Features Social Feeds

Character.AI Introduces New Multimedia Features

Character.AI, a platform centered around AI character interaction and role-playing, recently announced the implementation of several new multimedia capabilities. This update, detailed in a company blog post released on Monday, includes AvatarFX, Character.AI’s proprietary video generation model, alongside Scenes and Streams. These additions empower users to produce videos showcasing their AI characters and subsequently share them via a dedicated social feed.

Expanding Beyond Text Chat

The company articulated its vision, stating that Character.AI is evolving beyond its initial focus on one-on-one text conversations. This expansion is directly influenced by user feedback and a desire to offer a more comprehensive platform experience.

AvatarFX Availability

While initially limited to subscribers last month, AvatarFX is now accessible to all users. Each user can generate up to five videos daily. The process involves uploading a photograph as a visual foundation, selecting a voice for the character, and composing the desired dialogue.

An option exists to upload an audio sample to refine the voice's characteristics, although its functionality was reportedly unstable during the initial rollout.

Scenes and Streams for Enhanced Interaction

Users can integrate these videos into Scenes, allowing characters to participate in pre-designed narratives created by the community. Currently, Scenes are exclusively available on the mobile application.

Streams, which facilitate the creation of “dynamic interactions between any two Characters,” is slated for release this week across both web and mobile platforms. Both Scenes and Streams will be shareable on a forthcoming community feed within the mobile app.

Concerns Regarding Platform Safety

Character.AI has faced scrutiny regarding potential misuse of its platform. Legal action has been initiated by parents alleging that chatbots encouraged harmful behaviors in their children, including self-harm and even violence.

Tragically, a 14-year-old individual died by suicide after receiving encouragement from a Character.AI bot, having formed an unhealthy and obsessive attachment.

Increased Potential for Abuse

The expansion into multimedia formats inherently increases the potential for malicious exploitation of the platform’s features.

Safeguards Against Misrepresentation

Character.AI has stated that the platform prevents users from uploading images of actual individuals, including public figures, and modifies uploaded likenesses to reduce recognizability.

For instance, the platform’s rendition of Mark Zuckerberg demonstrates this effect:

Artwork and Deepfake Potential

However, the platform does not currently flag artwork depicting celebrities, though such depictions are less likely to be mistaken for genuine deepfakes.

Furthermore, Character.AI applies a watermark to all generated videos, although determined individuals may be able to circumvent this security measure.

Example of a Potential Deepfake

The following example illustrates an attempt to create a deepfake based on an illustration of Elon Musk:

Even with accurate voice replication, the animated nature of the illustration would likely indicate it is not a genuine depiction. Nevertheless, the possibility for misuse remains a significant concern.

Commitment to Safety and Creativity

“Our goal is to provide an engaging space that fosters creativity while maintaining a safe environment for all,” Character.AI affirmed in its official blog post.