Hey HN, we're the co-founders of LemonSlice (try our HN playground here: https://lemonslice.com/hn). We train interactive avatar video models. Our API lets you upload a photo and immediately jump into a FaceTime-style call with that character. Here's a demo: https://www.loom.com/share/941577113141418e80d2834c83a5a0a9Chatbots are everywhere and voice AI has taken off, but we believe video avatars will be the most common form factor for conversational AI. Most people would rather watch something than read it. The problem is that generating video in real-time is hard, and overcoming the uncanny valley is even harder.
We haven’t broken the uncanny valley yet. Nobody has. But we’re getting close and our photorealistic avatars are currently best-in-class (judge for yourself: https://lemonslice.com/try/taylor). Plus, we're the only avatar model that can do animals and heavily stylized cartoons. Try it: https://lemonslice.com/try/alien. Warning! Talking to this little guy may improve your mood.
Today we're releasing our new model* - Lemon Slice 2, a 20B-parameter diffusion transformer that generates infinite-length video at 20fps on a single GPU - and opening up our API.
How did we get a video diffusion model to run in real-time? There was no single trick, just a lot of them stacked together. The first big change was making our model causal. Standard video diffusion models are bidirectional (they look at frames both before and after the current one), which means you can't stream.
From there it was about fitting everything on one GPU. We switched from full to sliding window attention, which killed our memory bottleneck. We distilled from 40 denoising steps down to just a few - quality degraded less than we feared, especially after using GAN-based distillation (though tuning that adversarial loss to avoid mode collapse was its own adventure).
And the rest was inference work: modifying RoPE from complex to real (this one was cool!), precision tuning, fusing kernels, a special rolling KV cache, lots of other caching, and more. We kept shaving off milliseconds wherever we could and eventually got to real-time.
We set up a guest playground for HN so you can create and talk to characters without logging in: https://lemonslice.com/hn. For those who want to build with our API (we have a new LiveKit integration that we’re pumped about!), grab a coupon code in the HN playground for your first Pro month free ($100 value). See the docs: https://lemonslice.com/docs. Pricing is usage-based at $0.12-0.20/min for video generation.
Looking forward to your feedback!
EDIT: Tell us what characters you want to see in the comments and we can make them for you to talk to (e.g. Max Headroom)
*We did a Show HN last year for our V1 model: https://news.ycombinator.com/item?id=43785044. It was technically impressive but so bad compared to what we have today.
▲peddling-brink22 minutes ago
[-] I got really excited when I saw that you were releasing your model.
> Today we're releasing our new model* - Lemon Slice 2, a 20B-parameter diffusion transformer that generates infinite-length video at 20fps on a single GPU - and opening up our API.
But after digging around for a while, searching for a huggingface link, I’m guessing this was just a unfortunate turn of phrase, and you are not in fact, releasing an open weights model that people can run themselves?
Oh well, this looks very cool regardless and congratulations on the release.
Thanks! And sorry! I can see how our wording there could be misconstrued. With a real-time model, the streaming infrastructure matters almost as much as the weights themselves. It will be interesting to see how easily they can be commoditized in the future.
reply▲sid-the-kid7 minutes ago
[-] Thank you! We are considering to release an open-source version of the model. Somebody will do it soon. Might as well be us. We are mostly concerned with the additional overhead of releasing and then supporting it. So, TBD.
reply▲One thing I've learnt from movie production is actually what separates professional from amateur quality is in the audio itself. Have you thought about implementing personaplex from NVDIA or other voice models that can both talk and listen at the same time?
Currently the conversation still feels too STT-LLM-TTS that I think a lot of the voice agents suffer from (Seems like only Sesame and NVDIA so far have nailed the natural conversation flow). Still, crazy good work train your own diffusion models, I remember taking a look at the latest literature on diffusion and was mind blown by the advances in last years or so since u-net architecture days.
EDIT: I see that the primary focus is on video generation not audio.
This is a good point on audio. Our main priority so far has been reducing latency. In service of that, we were deep in the process of integrating Hume's two-way S2S voice model instead of ElevenLabs. But then we realized that ElevenLabs had made their STT-LLM-TTS pipeline way faster in the past month and left it at that. See our measurements here (they're super interesting):
https://docs.google.com/presentation/d/18kq2JKAsSahJ6yn5IJ9g...But, to your point, there are many benefits of two-way S2S voice beyond just speed.
Using our LiveKit integration you can use LemonSlice with any voice provider you like. The current S2S providers LiveKit offers include OpenAI, Gemini, and Grok and I'm sure they'll add Personaplex soon.
Thanks for sharing! Makes sense to go with latency first.
reply▲convivialdingo4 hours ago
[-] That's super impressive! Definitely one of the best quality conversational agents I've tried syncing A/V and response times.
The text processing is running Qwen / Alibaba?
Qwen is the default but you can pick any LLM in the web app (though not the HN playground)
reply▲sid-the-kid4 hours ago
[-] Thank you! Yes, right now we are using Qwen for the LLM. They also released a super fast TTS model that we have not tried yet, which is supposed to be very fast.
reply▲Wow this team is non-stop!!! Wild that this small crew is dropping hit after hit. Is there an open polymarket on who acquires them?
reply▲haha thank you so much! The team is incredible - small but mighty
reply▲Pricing is confusing
Video Agents
Unlimited agents
Up to 3 concurrent calls
Creative Studio
1min long videos
Up to 3 concurrent generations
Does that mean I can have a total of 1 minute of video calls? Or video calls can only be 1 minute long? Or does it mean I can have unlimited calls, 3 calls at a time all month long?
Can I have different avatars or only the same avatar x 3?
Can I record the avatar and make videos and post on social media?
Sorry about the confusion. Video Agents and Creative Studio are two entirely different products. Video Agents = interactive video. Creative Studio = make a video and download it. If you're interested in real-time video calls, then Video Agents is the only pricing and feature set you should look at.
reply▲thedangler16 minutes ago
[-] What happens if I want to make the video on the fly and save that to reuse it when the same question or topic comes up. No need to render a video. Just play the existing one.
reply▲You could add a Max Headroom to the hn link. You might reach real time by interspersing freeze frames, duplicates, or static.
reply▲sid-the-kid4 hours ago
[-] 1) yes on Max Headroom. we are on it.
2) it already is real time...?
reply▲Whoops! Mistook the "You're about to speak with an AI."-progress bar for processing delay.
reply▲I wonder if we should make the UI a more common interface (e.g. "the call is ringing") to avoid this confusion?
It's a normal mp4 video that's looping initially (the "welcome message") and then as soon as you send the bot a message, we connect you to a GPU and the call becomes interactive. Connecting to the GPU takes about 10s.
sid-the-kid3 hours ago
[-] Makes sense. The init should be about 10s. But, after that, it should be real time. TBH, this is probably a common confusion. So thanks for calling it out.
reply▲Wow I can’t get enough of this site! This is literally all I’ve been playing with for like half an hour. Even moved a meeting!
My mind is blown! It feels like the first time I used my microphone to chat with ai
This comment made my day! So happy you're liking it
reply▲sid-the-kid4 hours ago
[-] glad we found somebody who likes it as much as us! BTW, biggest thing we are working to improve is speed of the response. I think we can make that much faster.
reply▲We're launching a new AI assistant and I wanted to make it alive so I started to play around with LemonSlice and I loved it!! I wanted to make our assistant be like a coworker that can give it an ability to create Loom style videos. Here's what I created -
https://drive.google.com/file/d/1nIpEvNkuXA0jeZVjHC8OjuJlT-3...Anyway, big thumbs up for the LemonSlice team, I'm excited to see it progress. I can definitely see products start coming alive with tools like this.
How did your token spend add up? I'm hesitant with evil customers racking up ai charges just to shit and giggles. Even competitors might sponsor some runaway charges.
reply▲sid-the-kid5 hours ago
[-] Very cool! Thanks for sharing. I love your use-case of turning an AI coding agent into more of an AI employee. Will be interesting to see if users can connect better with the product this way.
reply▲dreamdeadline5 hours ago
[-] Cool! Do you plan to expose controls over the avatar’s movement, facial expressions, or emotional reactions so users can fine-tune interactions?
reply▲Yes we do! Within the web app, there's a "action text prompt" section that allows you to control the overall actions of the character (e.g. "a fox talking with lots of arm motions"). We'll soon expose this in the API so you can control the characters movements dynamically (e.g. "now wave your hand")
reply▲sid-the-kid5 hours ago
[-] Our text control is good, especially for emotions. For example, you can add the text prompt: "a person talking. they are angry", and agent will have an angry expression.
You can also control background motions (like ocean waves, or a waterfall or car driving).
We are actively training a model that has better text control over hand motions.
Heads up, your privacy policy[0] does not work in dark mode - I was going to comment saying it made no sense, then I highlighted the page and more text appeared :)
[0] https://lemonslice.com/privacy
sid-the-kid4 hours ago
[-] Fix deployed! This is why it's good to launch on hacker news. thanks for the tip.
reply▲koakuma-chan4 hours ago
[-] > You're probably thinking, how is this useful
I was thinking why the quality is so poor.
sid-the-kid4 hours ago
[-] curious what avatar you think is poor quality? Or, what you think is poor quality. i want to know :)
reply▲koakuma-chan4 hours ago
[-] Low res and low fps. Not sure if lipsync is poor, or if low fps makes it look poor. Voice sounds low quality, as if recorded on a bad mic, and doesn't feel like it matches the avatar.
reply▲sid-the-kid4 hours ago
[-] thanks for the feedback. that's helpful. Ya, some avatars have worse lip synch than others. It depends a little on how zoomed in you are.
I am double checking now to make 100% sure we return the original audio (and not the encoded/decoded audio).
We are working on high-res.
That's an interesting insight about "stacking tricks" together. I'm curious where you found that approach hit limits. And what gives you an advantage if anything against others copying it. Getting real-time streaming with a 20B parameter diffusion model and 20fps on a single GPU seems objectively impressive. It's hard to resist just saying "wow" looking at the demo, but I know that's not helpful here. It is clearly a substantial technical achievement and I'm sure lots of other folks here would be interested in the limits with the approach and how generalizable it is.
reply▲sid-the-kid2 hours ago
[-] Good question! Software gets democratized so fast that I am sure others will implement similar approaches soon. And, to be clear, some of our "speed upgrades" are pieced together from recent DiT papers. I do think getting everything running on a single GPU at this resolution and speed is totally new (as far as i have seen).
I think people will just copy it, and we just need to continue moving as fast as we can. I do think that a bit of a revolution is happening right now in real-time video diffusion models. There are so many great papers being published in that area in the last 6 months. My guess is that many DiT models will be real time within 1 year.
sid-the-kid2 hours ago
[-] One thing that is interesting: LLMs pipelines have been highly optimize for speed (since speed is directly related to cost for companies). That is just not true for real-time DiTs. So, there is still lots of low hanging fruit for how we (and others) can make things faster and better.
reply▲storystarling1 hour ago
[-] Curious about the memory bandwidth constraints here. 20B parameters at 20fps seems like it would saturate the bandwidth of a single GPU unless you are running int4. I assume this requires an H100?
reply▲Yep, the model is running on Hopper architecture. Anything less was not sufficient in our experiments.
reply▲sid-the-kid5 hours ago
[-] hey HN! one of the founders here. as of today, we are seeing informational avatars + roleplaying for training as the most common use cases. The roleplaying use-case was surprising to us. Think a nurse training to triage with AI patients. Or, SDRs practicing lead qualification with different kinds of clients.
reply▲Where’s the hn playground to grab a free month?
I have so many websites that would do well with this!
(We've replaced the link to their homepage (
https://lemonslice.com/) with the HN playground at the start of the text above)
Thanks Dan! The HN playground let's anyone try out for free without login
reply▲Wow this is the most impressive thing I’ve seen on hacker news in years!!!!!
Take my money!!!!!!
Wow thank you so much :) We're so proud of it!!
reply▲Wow this is really cool, haven't seen real-time video generation that is this impressive yet!
reply▲Thank you so much! It's been a lot of fun to build
reply▲I'm curious if I can plug in my own OpenAI realtime voice agents into this.
reply▲I'm using an openAI realtime voice with livekit, and they said they have a livekit integration so it would probably be doable that way. I haven't used video in livekit though and I don't know how the plugins are setup for it
reply▲Yes this is exactly right. Using the LiveKit integration you can add LemonSlice as an avatar layer on top of any voice provider
reply▲sid-the-kid5 hours ago
[-] Good question. When using the API, you can bring any voice agent (or LLM). Our API takes in what the agent will say, and then streams back the video of the agent saying it.
For the fully hosted version, we are currently partnered with ElevenLabs.
The last year vs this year is crazy
reply▲Agreed. We were so excited about the results last year and they are SO BAD now by comparison. Hopefully we'll say the same thing again in the couple months
reply▲sid-the-kid4 hours ago
[-] thanks! it just barley worked last year, but not much else. this year it's actually good. we got lucky: it's both new tech and turned out to be good quality.
reply▲Not working on mobile iOS
reply▲what's not working for you?
reply▲This looks super awesome!
reply▲sid-the-kid5 hours ago
[-] thank you! it's by far the thing I have worked on that I am most proud of.
reply▲This is next-level!
reply▲Thanks so much! We're super proud of it
reply▲ProjectBarks2 hours ago
[-] Removing - Realized I made a mistake
reply▲I don't see any evidence that r0fl's comments are astroturfing. Sometimes people are just enthusiastic.
I appreciate your concern for the quality of the site - that fact that the community here cares so much about protecting it is the main reason why it continues to survive. Still, it's against HN's rules to post like you did here. Could you please review https://news.ycombinator.com/newsguidelines.html? Note this part:
"Please don't post insinuations about astroturfing, shilling, bots, brigading, foreign agents and the like. It degrades discussion and is usually mistaken. If you're worried about abuse, email hn@ycombinator.com and we'll look at the data."
sid-the-kid2 hours ago
[-] it's a fair concern. but, we don't know r0fl. and we are not astroturfing.
even I am surprised with how many opnely positive comments we are getting. it's not been our experience in the past.