The human computing company.
Talk to a PAL today š£ļø
- Tweets 1,419
- Following 438
- Followers 11,226
- Likes 2,750
Introducing Phoenix-4.5, the fastest and most expressive real-time human rendering model on the market.
More natural than ever before: richer facial animation, more expressive emotion and micro-movements, and movement that now extends through the upper body.
It is the closest AI has ever come to passing the Turing test face to face.
98
90
683
182,035
Tavus retweeted
Gave GPT a face and feelings, called it "AGI Mode," then told it Anthropic's models were better.
Probs not surviving the singularity... so what should I ask next?
This is @OpenAIDevs GPT-Live-1 voice + @tavus Phoenix-4.5.
11
5
29
2,317
Most AI forgets you when the conversation ends. Ours doesn't.
Tavus PALs hit 93.4% on LongMemEval, the AI memory benchmark. That matches the best memory providers, and ours runs in real time. We published all 500 answers so you can check our work.
tavus.io/blog/memories
PALs now remember the people they talk to. Introducing Memories: memory built around the relationship between a PAL and each person it knows.
The art of a good conversation is everything you don't have to say twice and every conversation should build on the last.
9
3
32
1,437
Time is the one thing we can't render. So we're giving you twice as much of it.
Sign up for Tavus or upgrade your plan by the end of September to get 2Ć the minutes to build with.
maker.tavus.io
2
17
832
Packed house at our SF office for Audio Layer 3.0.
170+ builders, live demos (incl. @lightberry's Lumi), and a panel with @ai_coustics, @LiveKit + @GradiumAI on making real-time AI thrive in any scenario.
Our takeaway: there's never been a better time to build interactive AI.
2
15
745
Read the full post and benchmark breakdown: tavus.io/blog/memories
2
265
Tavus retweeted
Thereās so much conversation about what AI might do someday. Iām much more excited about the positive impacts it's having right now.
Careflick is using elderly care companions powered by Tavus PALs to fight senior loneliness. It's giving seniors the freedom of aging in place.
This is what positive real-world impact of AI looks like ā¤ļø
"Many of the seniors we serve go four or five days without seeing another face."
Careflick built AI companions for those days and Tavus gave them a face that talks.
97% of seniors chose video over voice-only, and 54% are still active at day 30.
tavus.io/customers/careflickā¦
3
2
13
1,138
"Many of the seniors we serve go four or five days without seeing another face."
Careflick built AI companions for those days and Tavus gave them a face that talks.
97% of seniors chose video over voice-only, and 54% are still active at day 30.
tavus.io/customers/careflickā¦
3
2
24
2,562
Tavus retweeted
Since no one seems to provide specifics on how AI is going to destroy us all, I decided to ask my AI PALs exactly how theyād go about it.
Can confirm: theyāre up to no good. Perfect alignment. Terrible governance.
For legal reasons, this is a joke. For scientific reasons, AI is not going to kill us.
Introducing Phoenix-4.5, the fastest and most expressive real-time human rendering model on the market.
More natural than ever before: richer facial animation, more expressive emotion and micro-movements, and movement that now extends through the upper body.
It is the closest AI has ever come to passing the Turing test face to face.
4
6
29
2,047
Tavus retweeted
Seeing some pretty incredible things since the 4.5 launch yesterday- the benchmarks and quality are great, but what Iām most excited about is that you can just feel the difference.
Here's one of our old faces versus 4.5 š
And the pace at which the research team is moving at to achieve these jumps is remarkable- Phoenix-4 has only been out for 5 months
Introducing Phoenix-4.5, the fastest and most expressive real-time human rendering model on the market.
More natural than ever before: richer facial animation, more expressive emotion and micro-movements, and movement that now extends through the upper body.
It is the closest AI has ever come to passing the Turing test face to face.
2
3
26
1,522
A PAL no longer has to be human.
Now with Phoenix-4.5, a single image is enough to turn a character into a PAL, whether anime, cartoon, or Pixar-style. It stays true to the original image, and it talks, listens, and moves in realtime like any other PAL.
maker.tavus.io
19
10
94
9,334
Phoenix-4.5 just launched yesterday.
It has the lowest latency in the world at 134 ms from audio to video, carries expression through the full upper body, and can create a talking PAL from just a single image.
See the full breakdown:
Introducing Phoenix-4.5, the fastest and most expressive real-time human rendering model on the market.
More natural than ever before: richer facial animation, more expressive emotion and micro-movements, and movement that now extends through the upper body.
It is the closest AI has ever come to passing the Turing test face to face.
6
885
Now imagine a PAL that can talk with its hands, stand up, and walk across a room (or even sing).
What a model that would be...
Introducing Phoenix-4.5, the fastest and most expressive real-time human rendering model on the market.
More natural than ever before: richer facial animation, more expressive emotion and micro-movements, and movement that now extends through the upper body.
It is the closest AI has ever come to passing the Turing test face to face.
12
5
52
3,265
Tavus retweeted
Hassaan-bench: Every realtime rendering model (including previous versions of Phoenix) chopped my face up badly. Phoenix-4.5 is the first time I don't cringe talking to my digital twin.
It's natural, expressive and makes me look so happy :)
Introducing Phoenix-4.5, the fastest and most expressive real-time human rendering model on the market.
More natural than ever before: richer facial animation, more expressive emotion and micro-movements, and movement that now extends through the upper body.
It is the closest AI has ever come to passing the Turing test face to face.
3
4
37
3,692
Introducing Phoenix-4.5, the fastest and most expressive real-time human rendering model on the market.
More natural than ever before: richer facial animation, more expressive emotion and micro-movements, and movement that now extends through the upper body.
It is the closest AI has ever come to passing the Turing test face to face.
98
90
683
182,035
And it shows up in the numbers:
ā¢ 134ms from audio to video (25% faster than any other model on the market)
ā¢Ā Lip sync improved 16%
ā¢Ā 64% of the faces that failed on Phoenix-4 now train successfully.
Read the research: tavus.io/blog/phoenix-4-5
Try it for yourself: maker.tavus.io
7
37
3,259
And it shows up in the numbers: 134 ms from audio to video (25% faster than any other model), lip sync improved 16%, and 64% of the faces that failed on Phoenix-4 now train successfully
Try it for yourself: maker.tavus.io
Read the research: tavus.io/blog/phoenix-4-5
1
2
155
