Lemon Slice, a pioneering digital avatar generation company, has successfully raised $10.5 million in seed funding to advance its innovative AI technology. The startup aims to revolutionize AI chatbots by integrating a dynamic video layer, enabling the creation of realistic digital avatars from just a single image. This funding round, led by Matrix Partners and Y Combinator, will fuel the development of its Lemon Slice-2 diffusion model, designed to transform text-based AI interactions into engaging visual experiences.
The seed funding round saw participation from notable investors including Matrix Partners, Y Combinator, Dropbox CTO Arash Ferdowsi, Twitch CEO Emmett Shear, and The Chainsmokers.
The core of Lemon Slice's offering is its Lemon Slice-2 model, a 20-billion-parameter diffusion model capable of live-streaming videos at 20 frames per second using a single GPU. This powerful technology allows for the generation of diverse digital avatars, from human-like figures to non-human characters, catering to various applications. Companies can integrate this model into their platforms via an API or an embeddable widget with minimal coding. Once an avatar is created, users retain full control over its background, styling, and overall appearance. For authentic voice generation, Lemon Slice leverages technology from ElevenLabs.
Co-founder Lina Colucci highlighted the company's vision for interactive video, stating,
“In the early days of GenAI, my co-founders started to play around with different video models, and it became obvious to us that video was going to be interactive. The compelling part about tools like ChatGPT was that they were interactive, and we want video to have that layer.”Colucci also critiqued current avatar solutions, noting,
“The existing avatar solutions I’ve seen to date add negative value to the product. They are creepy, and they are stiff. They look good for a few seconds, and as soon as you start interacting with them, it feels very uncanny, and it doesn’t put you at ease. The thing that has prevented avatars from really taking off is that they haven’t been good enough.”Lemon Slice aims to overcome this 'uncanny valley' effect by delivering highly realistic and engaging avatars.
Founded in 2024 by Lina Colucci, Sidney Primas, and Andrew Weitz, Lemon Slice believes its proprietary general-purpose diffusion model will differentiate it from competitors. This generative model learns to create new data by working backward from noisy training data, promising a superior approach to avatar generation.
Lemon Slice emphasizes its commitment to responsible AI development, implementing guardrails to prevent unauthorized face or voice cloning and utilizing large language models for content moderation. While specific clients remain undisclosed, the technology is already being deployed across various sectors, including education, language learning, e-commerce, and corporate training.
Lemon Slice enters a competitive market, facing established players in video generation like D-ID, HeyGen, and Synthesia, alongside other digital avatar creators such as Genies, Soul Machine, Praktika, and AvatarOS. However, investors express strong confidence in Lemon Slice's unique approach. Ilya Sukhar, a partner at Matrix, highlighted the team's technical strength and the broad applicability of their technology.
“It’s a deeply technical team with a track record of shipping ML products, not just demos and research. Many of the other players are bespoke to particular scenarios or verticals, and Lemon Slice is taking the generalized ‘bitter lesson’ scaling approach (of data and compute) that has worked in other AI modalities,”Sukhar commented. Jared Friedman of Y Combinator further emphasized the company's potential, noting that its diffusion-style model, similar to those used by Veo3 or Sora, positions it to achieve unprecedented realism.
“Lemon Slice is, I believe, the only company taking the fundamental ML approach that can eventually overcome the uncanny valley and break the avatar Turing test. They train the same type of model as Veo3 or Sora: a video diffusion transformer. Because it is a general-purpose model that does the whole thing end-to-end, it has no ceiling on how good it can get; the others top out below photorealistic. It also works for both human and non-human faces and requires only an image to add a new face,”Friedman explained.
With a current team of eight, Lemon Slice plans to strategically deploy the new capital to expand its engineering and go-to-market teams, alongside covering the significant compute costs associated with training its advanced AI models.







