Developers and companies are increasingly deploying AI agents and chatbots within their apps, but so far they’ve mostly been restricted to text. Digital avatar generation company Lemon Slice is working to add a video layer to those chats with a new diffusion model that can create digital avatars from a single image.
Called Lemon Slice-2, the model can create a digital avatar that works on top of a knowledge base to play any role required of the AI agent, like addressing customer queries, helping with homework questions, or even working as a mental health support agent.
“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,” co-founder Lina Colucci said.
Lemon Slice says this is a 20-billion-parameter model that can work on a single GPU to livestream videos at 20 frames per second. The company is making the model available through an API and an embeddable widget that companies can integrate into their sites with a single line of code. After an avatar is created, you can change the background, styling, and appearance of a character at any point.
Besides human-like avatars, the company is also focusing on being able to generate non-human characters to suit different needs. The startup is using ElevenLabs’ tech to generate the voices of these avatars.
Founded by Lina Colucci, Sidney Primas, and Andrew Weitz in 2024, Lemon Slice is betting that using its own general-purpose diffusion model (a type of generative model that learns to work backwards from noisy training data to generate new data) for making avatars will set it apart from competitors.
“The existing avatar solutions I’ve seen to date add negative value to the product,” Colucci said. “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.”
To fund that effort, the company on Tuesday said it has raised $10.5 million in seed funding from Matrix Partners, Y Combinator, Dropbox CTO Arash Ferdowsi, Twitch CEO Emmett Shear, and The Chainsmokers.
The company says it has guardrails in place to prevent unauthorized face or voice cloning, and that it uses large language models for content moderation.
Lemon Slice would not name the organizations using its technology, but said the model is being put to work for use cases like education, language learning, e-commerce, and corporate training.
The startup faces stiff competition from video generation startups like D-ID, HeyGen, and Synthesia, as well as other digital avatar makers Genies, Soul Machine, Praktika, and AvatarOS.
Ilya Sukhar, a partner at Matrix, thinks that avatars will be useful in areas where videos are prominent. For instance, people like learning from YouTube rather than reading long blocks of text. He noted that Lemon Slice’s technical prowess and its own will give it an edge over other startups.
“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,” he said.
Y Combinator’s Jared Friedman believes that using a diffusion-style model allows Lemon Slice to generate any kind of avatars as compared to some other startups that are focused on either human-like or game character-like avatars.
“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,” he said.
The startup currently has eight employees, and plans to use the funds to hire engineering and go-to-market staff, along with paying the compute bills to train its models.
