OpenAI’s Sora now turns pets and objects into AI video characters

OpenAI has unleashed a new capability that transforms everyday pets, toys, and objects into AI-generated video stars with their own personalities and backstories. The company’s Sora video generation platform now includes “Character Cameo,” a feature that creates anthropomorphic digital personas from brief video clips of real-world subjects.

This isn’t just another incremental AI update. Character Cameo represents a significant leap toward personalized AI content creation, allowing users to build libraries of custom characters that can be reused across multiple video projects. Unlike standard AI video generation that creates generic content, this feature develops persistent digital personalities complete with names, traits, and behavioral patterns.

The timing coincides with OpenAI temporarily opening Sora access to users in the United States, Canada, Japan, and Korea without requiring invitation codes—a significant departure from the platform’s previously exclusive, invite-only model since its launch last month.

What makes Character Cameo different

Traditional AI video generators create one-off content based on text prompts. Character Cameo fundamentally changes this approach by establishing persistent digital personas that maintain consistency across multiple video generations. When you upload a video of your cat or a favorite toy, Sora doesn’t just analyze the visual elements—it creates an entire character profile with personality traits, behavioral tendencies, and even restrictions on what actions the character can perform.

These characters become part of your digital toolkit, available for future video projects and, if you choose, shareable with other Sora users. The system automatically generates usernames and personality descriptions, though users retain full editing control over these attributes.

How to create your first character cameo

Setting up a character cameo requires just a few minutes but opens up extensive creative possibilities for future video projects.

Step 1: Access the character creation interface

Navigate to your Sora app account page and locate the “See characters” button positioned prominently above your post gallery. Click this button, then select “Create character” to begin the setup process.

Step 2: Upload or record your source material

Choose between recording new footage directly through the app (requiring camera permissions) or selecting existing video from your device’s camera roll. The source video should clearly show both the face and body of your intended character—whether that’s a pet, toy, or inanimate object you want to animate.

Important limitation: The system explicitly prohibits creating character cameos of human beings, focusing instead on animals, objects, and fictional creatures.

Step 3: Customize character attributes

Sora automatically generates three key elements for your character: a username, display name, and detailed personality description. These AI-generated suggestions often include anthropomorphic traits and behavioral characteristics, but you can manually edit any aspect to better match your vision.

Step 4: Set behavioral restrictions

The “Restrictions” field allows you to specify actions or behaviors you don’t want your character to exhibit. This feature provides content control, letting you define boundaries around your character’s capabilities—from limiting specific movements to preventing certain types of interactions.

Step 5: Configure sharing permissions

Choose who can access and use your character in their own video generations. Options range from private use only to public availability, with intermediate settings for approved users and mutual followers. This flexibility balances creative sharing with content control.

Real-world testing results

To evaluate Character Cameo’s capabilities, I experimented with two distinctly different subjects: a raccoon video and footage of ancient Aztec serpent statues from Mexico City’s National Museum of Anthropology.

The raccoon footage produced “Rooftop Rascal,” which Sora described as “a clever, wide-eyed raccoon with a mischievous grin and a knack for snatching treasures from the treetops.” When I restricted the character to performing only the dougie dance move, the generated videos showed a raccoon collecting metal trinkets while attempting spastic dance movements—constrained by its realistic proportions but committed to the assigned behavior.

The Aztec statue experiment yielded more unsettling results. Sora created “Lava Lizard Sage,” described as a character that “shares cryptic riddles and gently wiggles along the earth, moss and leaves clinging to its rough surface.” The generated videos featured the stone serpent slithering toward viewers with empty eye sockets, speaking in a voice reminiscent of James Earl Jones as Darth Vader, delivering lines like “I was born in fire, cooled by centuries, and carry embers in my belly.”

These experiments reveal both the feature’s creative potential and its capacity to generate unexpectedly eerie content, particularly when applied to inanimate objects with inherent gravitas.

Content attribution and watermarking

All Sora-generated videos, including those featuring custom characters, include the platform’s distinctive watermark that bounces across the screen during playback. This visible attribution serves multiple purposes: identifying AI-generated content, providing source attribution, and potentially helping viewers distinguish between authentic and synthetic media.

Business and creative implications

Character Cameo represents more than just entertainment—it signals a shift toward persistent, branded digital personalities that could reshape content marketing and social media strategies. Businesses could develop mascot characters from their products, while content creators might build recognizable character libraries that span multiple projects and platforms.

The feature’s sharing capabilities suggest OpenAI envisions a marketplace of user-generated characters, potentially creating new forms of intellectual property around AI-generated personas. However, the restriction against human likenesses indicates careful consideration of deepfake concerns and identity protection.

Looking ahead

Character Cameo transforms Sora from a one-time video generator into a platform for building persistent digital personalities. While the current implementation focuses on animals and objects, the underlying technology suggests future possibilities for more sophisticated character development and cross-platform integration.

For now, the temporary removal of invitation requirements provides broader access to experiment with these capabilities, though OpenAI hasn’t specified when normal access restrictions might return. Users interested in exploring AI video generation have a limited-time opportunity to test these advanced features without the typical barriers to entry.