'Runway Characters' Wants To Bring AI-Based Characters, But Responsibly

Runway recently introduced a new tool called 'Runway Characters,' a real-time video agent API designed to allow developers to create fully custom conversational characters.

According to the announcement, the system is built on Runway's GWM-1 world model and enables the generation of expressive digital personas from just a single reference image. These characters can appear in various styles, ranging from photorealistic humans to animated mascots or stylized brand figures, and they support natural facial expressions, eye movements, lip-syncing, and gestures both when speaking and listening.

The API provides extensive customization options, including control over voice, personality traits, embedded knowledge bases, and specific actions the character can perform during interactions.

No fine-tuning of models is required, and the system maintains consistent quality even during prolonged conversations.

Developers can integrate these characters into applications, websites, or other products, where they might handle tasks such as customer support, processing orders, creating tickets, or drawing from specialized information sources in real time.

For consumer access, preset avatars are available directly in the Runway web app, while enterprise teams can build custom implementations through the developer platform.

Potential applications include branded customer service experiences that align with a company's visual identity, interactive educational tools featuring tutors, historical figures, or fictional guides for extended learning sessions, and marketing scenarios with engaging animated mascots.

Runway has noted early adoption by partners such as BBC and Silverside, indicating practical deployments in professional contexts.

Alongside the launch, Runway published a follow-up post addressing the responsible development of such interactive AI characters.

The company acknowledges that technologies enabling highly immersive, real-time simulated interactions carry risks related to identity misrepresentation, deception, manipulation, and broader safety concerns.

They emphasize that powerful capabilities like generating avatars from a single image can support beneficial uses, such as consented educational recreations or enterprise support tools, but also require careful handling to prevent misuse, including non-consensual impersonation or harmful content generation.

To mitigate these issues at launch, Runway has implemented content and identity restrictions consistent with its usage policy.

These prohibit avatars depicting children, public figures without authorization, or protected intellectual property.

The system limits guidance on sensitive topics like medical, legal, or financial advice and avoids simulating therapeutic or counseling roles.

Users are encouraged to disclose that interactions involve AI-generated characters rather than real people, particularly in situations where the distinction might not be immediately clear.

A reporting mechanism exists for addressing unauthorized use of likenesses, and the company conducts ongoing red teaming to identify and address vulnerabilities before they are exploited.

Runway describes its approach as proactive rather than purely reactive, involving continuous monitoring, policy updates based on real-world feedback, and collaboration with industry partners, researchers, and policymakers.

They recognize that the technology is still evolving and that safeguards will need refinement over time, but the initial framework aims to balance accessibility and creative potential against foreseeable risks. This dual release of technical capability and accompanying responsibility discussion reflects an effort to introduce advanced AI interaction tools while openly considering their broader implications.