building realtime avatars @Anam__ai
London, England
Joined November 2024
- Tweets542
- Following113
- Followers855
- Likes789
Every conversation people have with our @Anam__AI avatars runs on @CoreWeave GPUs. They featured @Caoimhe_anam in their founder series, talking about how Anam got started + why we went with them
Like everyone else we’re GPU poor, so picking a provider matters a lot. We need the best GPUs as soon as they’re available, at a decent price, and across the US + Europe as we scale. Pretty fundamental when you’re training and running real-time video models like Cara-4.1 Express : )
More on how we work with them: go.anam.ai/coreweave-x
Ben Carr retweeted
An interactive avatar demo-ing itself. Who would've thought?! 🤠
Who better to explain Cara-4.1 Express than an avatar using Cara-4.1 Express
@snpranav got one of our @anam__AI avatars to walk through what's new: better lip sync, more expressive faces + director notes so you can control how it delivers each line
Live in the Lab now: go.anam.ai/c41/x-ben-lab
Who better to explain Cara-4.1 Express than an avatar using Cara-4.1 Express
@snpranav got one of our @anam__AI avatars to walk through what's new: better lip sync, more expressive faces + director notes so you can control how it delivers each line
Live in the Lab now: go.anam.ai/c41/x-ben-lab
Proud celts over here 🏴 🇮🇪
We also get asked ALL the time what @Anam__AI means + why we called our model Cara, so we share a little on that too @benatanam
Try it here: go.anam.ai/Anamlab-x
Across all six questions, 73% of Anam ratings were positive vs 54% for Tavus. 53/60 raters scored Anam higher than Tavus; 51/60 than HeyGen.
Median reply time: 1.24s vs HeyGen’s 1.82s.
Median connection time: 1.42s vs HeyGen’s 3.66s.
Anam’s visual-quality lead over Tavus was not statistically significant. Evaluation code and methodology are public, so you can reproduce the results.
Results: go.anam.ai/latencybenchmark
Repo: go.anam.ai/benchmark-repo
Ben Carr retweeted
@Anam_AI ran its new blind benchmark on a LiveKit pipeline, with the same speech recognition, LLM and voice for every provider, so the only thing that changed was the avatar.
If you're building a voice agent with LiveKit Agents, you can give it a face in a few lines of code.
Try it now: go.anam.ai/c41/partner-livek…
Cara-4.1 Express is now live at @Anam__AI
Faster and more expressive than cara-4 with better lip sync and stronger prompt adherence. It also handles much wider framings for custom avatars.
We ran blind human evals: 85% of raters scored Anam higher than HeyGen, 88% higher than Tavus and 92% higher than LemonSlice. The eval code is public and reproducible.
Try it in the Anam Lab: go.anam.ai/dXLpQWy
Ben Carr retweeted
AI avatars have looked realistic for a while.
The problem starts when you actually talk to them.
Anam's new Cara-4.1 Express responds in just 1.2 secs, handles interruptions, and makes eye contact feel natural.
This is getting surprisingly close to a real conversation.
🤝 Paid partnership
Cara-4.1 Express is now live at @Anam__AI
Faster and more expressive than cara-4 with better lip sync and stronger prompt adherence. It also handles much wider framings for custom avatars.
We ran blind human evals: 85% of raters scored Anam higher than HeyGen, 88% higher than Tavus and 92% higher than LemonSlice. The eval code is public and reproducible.
Try it in the Anam Lab: go.anam.ai/dXLpQWy
Ben Carr retweeted
AI avatars can look impressive in a demo.
Then you start talking to them, and the awkward pauses, expressions, or lip sync give them away.
That's what makes Cara-4.1 Express from @Anam__ai so interesting.
1.2s response time, better lip sync, and more natural reactions are aimed at making conversations feel less robotic.
The real test isn't how an avatar looks in a still frame.
It's how it feels in a conversation.
Cara-4.1 Express is now live at @Anam__AI
Faster and more expressive than cara-4 with better lip sync and stronger prompt adherence. It also handles much wider framings for custom avatars.
We ran blind human evals: 85% of raters scored Anam higher than HeyGen, 88% higher than Tavus and 92% higher than LemonSlice. The eval code is public and reproducible.
Try it in the Anam Lab: go.anam.ai/dXLpQWy
Ben Carr retweeted
YOU COULD GET 25%+ MORE CONVERSIONS FROM YOUR AI SALES AGENT JUST BY GIVING IT A FACE.
Cara-4.1 express from @Anam__AI is a real-time face builders can add to their agent:
→ 70% prefer it to voice-only
→ 57% higher retention
→ real-time conversations in 70+ languages
this could literally lift conversion across every funnel. we're definitely testing it in ours.
Cara-4.1 Express is now live at @Anam__AI
Faster and more expressive than cara-4 with better lip sync and stronger prompt adherence. It also handles much wider framings for custom avatars.
We ran blind human evals: 85% of raters scored Anam higher than HeyGen, 88% higher than Tavus and 92% higher than LemonSlice. The eval code is public and reproducible.
Try it in the Anam Lab: go.anam.ai/dXLpQWy
Ben Carr retweeted
Every AI avatar looks great for about 5 seconds.
Then you actually talk to it. dead eyes, lips a little off, that weird pause before every reply. the whole thing falls apart.
Anam just released Cara-4.1 Express and it’s the first one that looks like it’s trying to fix that part.
Conversational AI went text, then voice, now video. We trust a face more than a text box, and adding one to a text or voice agent drives a 44% increase in engagement.
What stands out to me:
– 1.2 seconds response time, 33% faster than the next best competitor
– you can interrupt it mid-sentence and it handles it like a real conversation
– 73% of ratings positive vs 54% for Tavus
– #1 overall in a blind benchmark with 60 raters and 960 conversations
Free to start: anam.ai
🤝 Paid partnership
Cara-4.1 Express is now live at @Anam__AI
Faster and more expressive than cara-4 with better lip sync and stronger prompt adherence. It also handles much wider framings for custom avatars.
We ran blind human evals: 85% of raters scored Anam higher than HeyGen, 88% higher than Tavus and 92% higher than LemonSlice. The eval code is public and reproducible.
Try it in the Anam Lab: go.anam.ai/dXLpQWy
Ben Carr retweeted
the thing that makes an AI avatar feel real isn't just how it looks, it's how naturally it responds
Anam’s new Cara-4.1 Express model has better expressions, smoother lip sync, and came out #1 again in another blind test against its competitors.
I'm curious to see how developers use this for AI tutors and interactive assistants. it’s human nature, students pay more attention to a face
#AnamPartner
🤝 Paid partnership
Cara-4.1 Express is now live at @Anam__AI
Faster and more expressive than cara-4 with better lip sync and stronger prompt adherence. It also handles much wider framings for custom avatars.
We ran blind human evals: 85% of raters scored Anam higher than HeyGen, 88% higher than Tavus and 92% higher than LemonSlice. The eval code is public and reproducible.
Try it in the Anam Lab: go.anam.ai/dXLpQWy
Cara-4.1 Express is now live at @Anam__AI
Faster and more expressive than cara-4 with better lip sync and stronger prompt adherence. It also handles much wider framings for custom avatars.
We ran blind human evals: 85% of raters scored Anam higher than HeyGen, 88% higher than Tavus and 92% higher than LemonSlice. The eval code is public and reproducible.
Try it in the Anam Lab: go.anam.ai/dXLpQWy
3 years of building @Anam__AI, and @Caoimhe_anam and we haven’t killed each other yet :)
Oct ‘23 we both quit our jobs to build the next interface for technology. The first few months were just the two of us trying to put together a deck, get a POC working, find our first customers and convince a few researchers and engineers to take a chance on us.
By May ‘24 we had money in the bank and our first 3 hires. Since then:
- #1 ranked real-time model, 3 releases in a row new models shipping every quarter
- 60K+ users signed up to try, buy and build on Anam
- XM+ monthly minutes in production over the last 4 weeks
- 14x ARR in the last 13 months
- a team of absolute superstars across GTM to research
- backed by some of the best VCs around
First pic is from hours of live testing our latest model in a dark room in the London office. Second is Caoimhe in my old apartment the night we got the very first POC working.
V cool to look back on, but we’re just getting started
