Google makes it easy to deepfake yourself
YouTube rolls out a new AI feature to generate realistic, 8-second avatar videos that look and sound like you.
Google is introducing a new AI avatar feature for YouTube Shorts, giving creators a controlled way to generate digital clones of themselves. To create an avatar, users must record a "live selfie" capturing their face and voice under specific conditions: good lighting, a quiet area, and a clean background. Once created, the avatar can generate new 8-second video clips from text prompts or be inserted into existing eligible Shorts. The company frames this as a "safer and more secure" method for AI content creation, contrasting it with the unregulated deepfake landscape.
The feature comes with significant restrictions and safeguards. Avatars can only be used in the creator's own original videos, and creators maintain full control, including the ability to delete their avatar or any videos featuring it at any time. All content generated with the tool will be clearly labeled as AI, using visible watermarks and digital authentication markers like Google's SynthID and the broader C2PA standard. The launch follows OpenAI's recent sunsetting of its Sora video generation platform, which faced copyright and deepfake controversies. This move expands YouTube's suite of creator tools, many powered by Google's Gemini models, and reflects the platform's ongoing effort to balance generative AI innovation with content safety.
- Creators generate an AI avatar by recording a live selfie with face and voice under specific conditions.
- The avatar can produce new 8-second videos from prompts and will be clearly watermarked with SynthID/C2PA labels.
- Tool rolls out gradually to creators 18+ with channels, positioning it as a safer alternative following OpenAI's Sora shutdown.
Why It Matters
It provides a controlled, labeled path for AI-generated content, addressing deepfake concerns while expanding creative tools for professionals.