We're excited to be the voice AI partner supporting Lemon Slice Live!
Lemon Slice is pioneering the next generation of expressive AI with a video foundation model that lets anyone create and interact with animated, talking characters. Their newest release, Lemon Slice Live, takes this vision even further. It allows users to video chat in real time with characters generated from any photo or illustration.
Whether it’s a photorealistic rendering of a historical figure or a stylized cartoon, Lemon Slice turns static images into fully animated, responsive personas. As their voice AI provider, Deepgram powers the real-time, lifelike voice interactions that bring these characters to life.
See Lemon Slice Live in action below:
What is Lemon Slice?
A Y Combinator-backed startup founded by AI researchers and creative technologists, Lemon Slice is building foundational models for expressive, multimodal interaction. Their platform is designed for creators, educators, developers, and brands looking to bring characters to life across storytelling, marketing, entertainment, and virtual engagement.
Lemon Slice stands out for its real-time animation and interactivity, powered by research in diffusion transformer models and a unified experience that combines visuals, voice, and personality.
Introducing Lemon Slice Live
Lemon Slice Live is the company’s most ambitious release to date. It is a real-time video chat interface that transforms any image, such as a photo, illustration, or painting, into a responsive, speaking character. Simply upload a picture and engage in a live conversation, complete with accurate lip sync, expressive facial animation, and dynamic voice responses.
Unlike other tools that require character-specific training or motion rigging, Lemon Slice Live uses a zero-shot video diffusion transformer model. It works out of the box without setup and supports 10 languages. The system streams video at 25 frames per second with strong temporal consistency, allowing characters to maintain visual realism even across extended conversations.
This is not just video generation; it is real-time embodied AI where latency, quality, and realism all matter.
Why Lemon Slice Chose Deepgram
To enable seamless, lifelike interactions, Lemon Slice needed a voice AI platform that could keep up with its high-speed video engine and support a fully streaming AI architecture. Their real-time system coordinates several layers — voice transcription, language model inference, text-to-speech synthesis, and video generation — all of which must operate without delay. That meant selecting a voice AI platform that delivers:
- Low latency for a real-time experience
- High accuracy to ensure natural, intelligent replies
- Developer-friendly APIs that are quick to integrate
Deepgram met all three requirements. It was the clear choice to power Lemon Slice’s voice layer. “We chose Deepgram as our partner because they offer the fastest and most reliable AI voice services available today,” said Andrew Weitz, Co-founder of Lemon Slice. “Their developer-friendly streaming API made it easy to quickly integrate into our pipeline without sacrificing speed.”
By integrating Deepgram’s real-time voice AI, Lemon Slice ensures that user speech is processed with minimal delay. This helps the system stay within its current end-to-end latency target and enables more fluid and believable character interactions.
A New Era for Interactive AI Experiences
Lemon Slice represents the kind of innovation we’re proud to support: fast-moving, voice-native, developer-led, and focused on building expressive applications that feel alive. This partnership shows what’s possible when Deepgram’s enterprise-grade voice infrastructure is combined with generative video models built for interactivity.
We’re honored to be part of Lemon Slice Live’s launch and excited to see how developers, educators, and creators use this technology to reshape how we interact with digital characters.
Check out Lemon Slice Live to experience it for yourself. If you’re building real-time voice experiences of your own, try out Deepgram’s Streaming Speech-to-Text and Aura-2 Text-to-Speech in our API Playground.