We’ve tested our Audio Decisions API on iconic performances. All kinds of iconic.
This rooftop scene swings from peak fury to casual pleasantries in seconds, and the model tracks every pivot. Emotion detection on hard mode.
Not a product, just a glimpse of what this model makes possible.
Our team’s so good at optimization, they even cut the time between performance breakthroughs.
Two weeks since launch, the Audio Decisions API is faster across the board. Responses now average ~160ms, with 95% back in about 200ms, and our slowest responses are down ~60%.
Try it yourself, no account required. 250,000 free tokens (~50 min of audio).
Hume Audio Decision Benchmarks hume.ai/blog/hume-audio-deci… via @hume_ai
1
26
one of my favorites
Last time, two people. This time, the room gets a lot more crowded. Research engineer Hoon Shin ran our Audio Decision API over another iconic dramatic scene to see how it keeps up with more voices.
Each orb is a line of dialogue. Its color shows the emotional mix, its size and pulse show intensity, and a red flare means someone’s raising their voice. With more voices in the scene, the threads have a lot to keep up with while the gauge tracks the tension.
All from the audio alone.
Not a product, just a glimpse of what this model makes possible.
1
33
Hume AI has been named to CB Insights and Accenture’s NYC Emerging AI 50, which recognizes the most promising AI startups based in New York.
We’re listed under AI infrastructure, and that’s exactly where we think voice understanding belongs: underneath the products, helping every system that listens to people go beyond the transcript and actually listen.
Thank you to @CBinsights and @Accenture for the recognition, and to the incredible team here whose work earned it.
Been demoing this all week and the moment it lands is always the same: someone asks a few questions about their own call recording and gets answers back with a confidence score on each.
If you work with audio, are building agents, or in the Voice AI space at all, this is worth a 30 second watch. Drop thoughts below or DM me with questions :)
The Hume Audio Decision API in 30 seconds, for anyone who missed the launch.
Ask it anything about a recording and it answers from the sound itself, not the transcript. A sneeze, a shift in tone, a change of accent. Answers in under a second.
One model for every question you have about audio.
1
1
66
Experimental demo: Scene Analysis
In one of his earliest explorations, research engineer Hoon Shin asked: what happens when our Audio Decision API listens to a scene from a movie?
Each orb is a line of dialogue. Its color shows the emotional mix, its size and pulse show intensity, and a red flare means someone’s raising their voice. Threads trace speaker changes, and the gauge tracks tension across the scene,
All from the audio alone.
Not a product, just a glimpse of what this model makes possible.
For those curious about batch mode: this uncut demo shows the Hume Audio Decision API working through hours of recorded calls, in real time.
Reading this post probably took longer than the processing did.
All your questions. All your audio. Confidence scores on every answer.
Full demo below.
Introducing the Hume Audio Decision API.
Ask anything: How many speakers? Is anyone laughing? Is the caller frustrated?
What used to take 10+ models is now one call.
Trained on 10M+ human judgments, so it answers the way human listeners would.
Realtime, or across hours of recordings.
Request access below.
andrew muggia retweeted
Our new Audio Decision model is INSANE. This is the fastest we've ever shipped since I joined and it might be our finest work. The research team at @hume_ai is at the top of their game.
Introducing the Hume Audio Decision API.
Ask anything: How many speakers? Is anyone laughing? Is the caller frustrated?
What used to take 10+ models is now one call.
Trained on 10M+ human judgments, so it answers the way human listeners would.
Realtime, or across hours of recordings.
Request access below.
Hume AI launches the Jev of voice AI thedeepview.com/articles/hum…
1
75
Ask your audio any questions, give it the possible answers, and get a confidence score for each one. Data you can actually act on. Incredibly powerful, and the use cases are endless. Incredibly proud of the entire @hume_ai team - Let's Go!!
Voice AI ̶h̶a̶s̶ had a listening problem.
It could transcribe every word and still miss what was actually happening.
Today we’re launching the Hume Audio Decision API, built on the world’s first audio-native decision model.
Ask it anything about the audio:
→ How many speakers?
→ Is anyone laughing?
→ Is the caller frustrated?
What used to take 10+ models is now one call.
Trained on 10M+ human judgments, so it hears the way people do.
Real time, or across hours of recordings.
Check out the below 📷 Watch it listen.
1
3
86
Hume AI just relaunched our Expression Measurement API with new models for voice and face.
Expression Measurement captures not only WHAT was said, but HOW it was said. It turns forms of expression into scores you can search, compare, and act on. DM me for more details.
2
100
Someone says "I'm fine." You know they aren't.
You heard it - the pause, the flat tone. Voice AI doesn't. It all gets dropped at transcription.
So we built the Expression Measurement API. 48+ emotions, 600+ descriptors, 50+ languages, audio and video, real-time.
Conversation was always more than words.
andrew muggia retweeted
Listen Labs is joining @Salesforce. We’ve entered into a definitive agreement for the acquisition and expect the transaction to close in the coming months.
Every day, people use Listen to hear from thousands of customers at once and turn what they say into decisions their teams can act on. They've shown what customer understanding looks like when asking customers is fast and easy.
Now we get to bring that to far more companies as part of Salesforce AI Labs.
Customers - you'll keep getting the Listen you know, and the people behind it stay the same.
Thank you to every researcher who ran a study with us, pushed us with feedback, and made Listen better along the way.
We're just getting started.
Text-to-speech was built to read a single voice aloud. Now it's being asked to hold a conversation between two, and that's a different problem entirely.
The speakers have to stay distinct, trade turns with natural timing, and keep their emotional states from bleeding into each other. None of that shows up in word error rate.
So we built a rubric: five dimensions, 48 two-speaker scripts, and four scenarios designed to break a dialogue by pairing similar-sounding voices, forcing frequent interruptions, contrasting emotions, and running long exchanges.
Three blind human raters per clip. We ran three generations of Google's Gemini TTS models through it, alongside other multi-speaker systems.
Keeping two voices apart is the hardest part of the job. Every system handles a male and female pair easily and drops on at least one same-gender pairing. The newest Flash model made its biggest gain on exactly that.
Models with the same overall score get there differently. One leads on keeping speakers’ voices distinct, another on making the dialogue sound like one continuous recording. The overall score hides that difference.
New capabilities need new evaluations.
Full case study, with audio examples, is linked in the comments.







