Character.AI’s AvatarFX Video Model Can Make Images Come to Life

The post Character.AI’s AvatarFX Video Model Can Make Images Come to Life appeared first on Android Headlines.

Apr 23, 2025 - 12:31

Summary: Character.AI has announced its first artificial intelligence-powered video generation model. Called AvatarFX, the platform is capable of generating high-quality videos from a pre-existing image. The company has put a handful of safety guards to prevent the misuse of the new model to create potentially harmful content or deepfakes.

Character.AI, one of the first artificial intelligence-powered platforms, introduced its first video generation model, AvatarFX. It is a cutting-edge AI model capable of creating photorealistic videos with audio. Character.AI’s AvatarFX AI video generation model distinguishes itself from other AI video models like OpenAI’s Sora by allowing users to turn static images into lifelike videos. In the platform’s own words, “AvatarFX can make images come to life – and speak, sing, and emote – all with a click of a button.”

Character.AI’s AvatarFX can create high-quality videos from pre-existing images

As per Character.AI, rather than relying on text prompts for video generation, AvatarFX allows users to use a pre-existing photo to create a high-quality video. It is capable of creating top-quality video of 2D animated characters, 3D cartoon characters, pets and pets, among others. The AI video model can maintain temporal consistency with face, hand, and body movement.

Furthermore, Character.AI mentions that AvatarFX can maintain that temporal consistency even in long-form videos. It means the subject of your video will remain consistent between frames, while glitches like extra arms and unwanted facial expressions will be less frequent. As for the architecture, Character.AI has built AvatarFX on flow-based diffusion models.

Character.AI said that thanks to the pre-existing image input prompts, users have better control over the generated output. Besides, the videos will also support speech, which the AI tool will generate using native TTS (text-to-speech) models.

What about safety?

While Character.AI’s AvatarFX video model sounds promising, it raises concerns about deepfakes and the generation of potentially harmful content. Character.AI said that it has taken multiple safety measures to mitigate the risk of deepfakes. The company said that it runs the dialogue entered by users through its safety filters, which flag content that would violate its policies.

Furthermore, the AI model doesn’t allow creating videos using photos of minors, high-profile politicians, and notable figures. The video generation process also renders other human images unrecognizable. AvatarFX also applies a watermark to the generated videos to distinguish them from real footage.

Character.AI also mentioned that it is setting new terms of usage for the new feature. The terms prohibit “impersonation, bullying, deepfakes, and use of protected IP without protection”, and anyone violating them would face a “strict” one-strike ban.

That said, the company said that AvatarFX is currently available in the testing period for CAI+ subscribers. However, there’s no word regarding its availability for the free tier at the moment.

The post Character.AI’s AvatarFX Video Model Can Make Images Come to Life appeared first on Android Headlines.