Lemon Slice Raises $10.5M to Advance Digital Avatar Tech
▼ Summary
– Lemon Slice has developed a new AI model, Lemon Slice-2, that creates interactive digital avatars from a single image to add a video layer to AI agents and chatbots.
– The 20-billion-parameter model can livestream video at 20 frames per second on a single GPU and is offered via an API for easy integration into applications.
– The startup differentiates itself by using its own general-purpose diffusion model to generate both human and non-human avatars, aiming to overcome the “uncanny valley” effect of previous solutions.
– The company has raised $10.5 million in seed funding to hire staff and fund compute costs, targeting use cases like education, e-commerce, and corporate training.
– It faces competition from other avatar and video generation startups, but investors believe its technical approach and scalable model give it a potential edge.
The landscape of AI interaction is poised for a significant visual upgrade, moving beyond static text to dynamic, conversational video. Lemon Slice, a digital avatar generation startup, has secured $10.5 million in seed funding to advance its mission of bringing realistic, interactive avatars to apps and websites. The company’s core technology, a diffusion model called Lemon Slice-2, aims to transform standard chatbots and AI agents by giving them a video presence that can be tailored for diverse roles, from customer service to educational support.
Unlike many current solutions that produce stiff or unsettling digital characters, Lemon Slice is building its platform on a proprietary, general-purpose diffusion model. This approach, similar to the architecture behind advanced video generation tools, allows the system to create avatars from a single image. The resulting characters can be human or non-human, and their backgrounds, styling, and appearance can be modified at any time. Co-founder Lina Colucci argues that previous avatar technologies have “added negative value” by failing to be genuinely engaging or putting users at ease, a gap her company intends to close.
The technical foundation is a 20-billion-parameter model designed to operate efficiently, capable of livestreaming video at 20 frames per second on a single GPU. For integration, Lemon Slice offers an API and an embeddable widget, allowing companies to add a video avatar to their platforms with minimal code. The startup partners with ElevenLabs to generate the voices for these avatars, creating a complete audiovisual package. Importantly, the company states it has implemented safeguards to prevent unauthorized face or voice cloning and uses large language models for content moderation.
Investors in the seed round include Matrix Partners, Y Combinator, and notable angels like Dropbox CTO Arash Ferdowsi and Twitch CEO Emmett Shear. The funding will support hiring in engineering and go-to-market roles, as well as covering the substantial compute costs required for model training. While Lemon Slice did not disclose specific clients, it highlighted target applications in education, language learning, e-commerce, and corporate training.
The market for digital avatars is crowded, with competitors ranging from video generation specialists like D-ID, HeyGen, and Synthesia to avatar-focused firms such as Genies and Soul Machine. However, Lemon Slice’s backers believe its fundamental technical strategy provides a distinct advantage. Ilya Sukhar of Matrix Partners praised the team’s “deeply technical” background and their commitment to a scalable, generalized model rather than bespoke solutions for narrow verticals.
Jared Friedman from Y Combinator offered a particularly bold assessment, suggesting Lemon Slice’s diffusion transformer model, aligned with the approach of leading video AI, is uniquely positioned to eventually “overcome the uncanny valley and break the avatar Turing test.” He emphasized that because the model is end-to-end and general-purpose, its potential quality is not artificially capped, allowing it to work seamlessly for any type of face or character. With eight employees currently, the startup is now focused on growing its team and refining its technology to make interactive video avatars a standard, and more natural, layer of digital communication.
(Source: TechCrunch)
