YouTube is rolling out an AI tool that lets creators build realistic digital clones of themselves for use in Shorts videos. The feature, reported by The Verge AI, lets users create what YouTube calls an “avatar” that will “look and sound like you” and can star in new or existing short-form content.
The timing is notable. Google is doubling down on generative video tools for creators at the exact moment OpenAI is retreating from the space, having shut down its Sora video platform last month after copyright headaches and a flood of low-quality content.
How It Works
The setup process is straightforward:
- Record a live selfie following on-screen prompts to capture your face and voice
- Use good lighting, a quiet room, no other faces in the background, and hold your phone at eye level
- Generate clips up to eight seconds long from text prompts, or insert your avatar into “eligible Shorts” in your feed
YouTube hasn’t clarified what makes a Short “eligible” for avatar insertion, which leaves a pretty big question mark over one of the feature’s core use cases.
The Guardrails
Google clearly knows this feature sits in uncomfortable territory. A tool that makes it trivially easy to create a photorealistic video clone of yourself is, by definition, deepfake technology. The company is framing it as a “safer and more secure” approach, with several restrictions:
- Avatars can only appear in the creator’s own original videos
- Creators control whether their Shorts can be remixed
- Avatars and generated videos can be deleted at any time
- Unused avatars get automatically deleted after three years
- All avatar content gets flagged as AI-generated with visible watermarks, SynthID, and C2PA labels
- Users must be at least 18 and own an existing YouTube channel
The C2PA labeling is worth a skeptical glance. As The Verge AI notes, it’s “broadly supported but questionably useful” as an authentication marker. Watermarks and metadata labels only work if downstream platforms and audiences actually check them.
Why This Matters
This is Google leaning hard into its position as the dominant player in AI-powered creator tools. YouTube’s AI suite now includes generated video clips, auto-dubbing, and an analytics chatbot, most of it powered by Gemini models.
The strategic picture is clear: Google wants creators producing more content, more often, with lower friction. An avatar that can film Shorts while you sleep (or just don’t feel like being on camera) fits that playbook perfectly.
But the platform is walking a tightrope. YouTube is still struggling to contain AI slop, deepfake scams, and impersonation content, according to The Verge AI. Releasing an official tool that generates photorealistic video clones, even with guardrails, adds complexity to that fight.
The feature is rolling out gradually with no specific timeline or regional availability announced. For more details, check the original coverage at The Verge AI.