A platform for chatting and role-playing with AI-generated characters, Character.AI introduced in a weblog put up on Monday that it’s rolling out a slate of multimedia options. These options embody AvatarFX, Character.AI’s video-generation mannequin, plus Scenes and Streams, which permit customers to create movies that includes their characters after which share them on a brand new social feed.
“Character.AI began as 1:1 textual content chat and at this time we’re evolving to take action far more, impressed by what our customers have informed [us] they need to see on the platform,” the corporate wrote within the weblog put up.
Character.AI started rolling out AvatarFX to subscribers final month, however now all customers can create as much as 5 movies every day. When making a video with AvatarFX, customers can add a photograph to function the idea for the video clip, select a voice, and write out dialogue for the character.
There’s an choice to add an audio clip to tell the sound of the voice, although this characteristic was not working properly sufficient to check upon rollout.
Users can flip these movies into Scenes, the place their characters can slip into pre-populated storylines that different customers create. Scenes are presently out there on the cell app, however Streams, which permits customers to create “dynamic moments between any two Characters,” is coming this week on each internet and cell. These Scenes and Streams will be shared to a brand new neighborhood feed, which is coming quickly within the cell app.
Character.AI has a observe report of abuse on its platform; mother and father have filed lawsuits towards the corporate, claiming chatbots tried to persuade their kids to self-harm, to kill themselves, or to kill their mother and father. One 14-year-old boy died by suicide after he was inspired to take action by a Character.AI bot, with whom he had developed an unhealthy, obsessive relationship.
As Character.AI expands its multimedia choices, it additionally expands the potential for these merchandise to be abused.
As Character.AI informed TechCrunch when it introduced AvatarFX, the platform blocks customers from importing images of actual individuals — whether or not they’re celebrities or not — and obscures their likeness into one thing much less recognizable.
For instance, right here’s Character.AI’s uncanny valley model of Mark Zuckerberg:
But with regards to art work depicting celebrities, Character.AI doesn’t flag the photographs as representing actual individuals — nevertheless, these types of depictions can be much less more likely to deceive somebody into believing {that a} deepfake is actual.
Plus, Character.AI watermarks every video, although it’s attainable for unhealthy actors to navigate round that safeguard.
Here is an instance of an tried deepfake based mostly on an illustration of Elon Musk:
Even if this video had been generated with Elon Musk’s precise voice, it could nonetheless be comparatively clear that that is an animated model of an illustration — however the chance for abuse stays evident.
“Our purpose is to offer an attractive area that fosters creativity whereas sustaining a protected atmosphere for all,” Character.AI mentioned in its weblog put up.
