ByteDance has suspended Seedance 2.0 photo-to voice feature after the days of the AI video generation model’s release that took the internet by storm.
Seedance 2.0 is capable of creating 1080 high resolution clips from simple text, images, audio, or video based prompts. Some users are calling it “the next evolution of AI video.”
In the wake of emerging technical controversies, ByteDance has decided to hit pause on the photo-to-voice tool amid privacy and safety concerns.
The model reportedly generates highly accurate personal voice features only by using facial images even without user authorization or consent.
The company issued a statement, “To maintain a healthy and sustainable creative environment, we are making urgent changes based on user feedback and will not allow real-human-like photos or videos to be used as reference subjects.”
The privacy concerns came to fore in a recent test performed by Pan Tianhong’s media outlet MediaStorm. The outlet discovered a personal photo could allow the AI model to produce a voice nearly identical to the original one without using any voice-based prompt.
Given the industrialization of deepfake fraud and AI-enabled scams, such revelation has raised public concern over AI regulation. If misused, the feature could be exploited to enable new forms of online abuse.
Moreover, ByteDance’s Jimeng and Doubao apps introduced a live verification step. The users will be obligated to record their own image and voice developing a digital avatar.
Such measures will not only mitigate privacy risks but also ensure regulatory compliance without compromising innovation.