iharnoor

is it pre -trained form scratch and our based on OSS model? how do you evaluate btw

love the idea, how many languages can you support?

kinda makes sense how Nvidia gets more distribution after acquiring hugging face, right?

i believe OSS STT is amazing but htings break udring mulaw and turn detection where closed source models still are ahead

Did you measure cold-start latency and peak memory on the phone builds as well as steady-state audio throughput?

why not a screenshot app? with arrows

why not just use chatgpt voice?

it comes with the analogy of art, Art is something that comes with human emotion, not something that's just beautiful enough

isn't harness as important as model these days?

I am hyped to use Jev with AssemblyAI for voice agents and TTS but you can't rely on Jev for high reasoning tasks at all.

for reasoning it less intelligent model we need fallbacks to more intelligent models with good routing

also we can't rely on Jev for things that require more thinking as Jev is not intelligent as Astra or Opus

curious, what inference providers your team is using and does that change the questions for ZDR etc?

Did you try considering any alternative models as per Soniox's STT benchmark?

Today, Universal-3.5 Pro is available on OpenRouter, served through our Sync Speech-to-Text API.

If you're already routing your LLM calls through OpenRouter, the speech-to-text call at the front of your pipeline now runs on the same key.

50% off list through OpenRouter for the first week: https://www.assemblyai.com/blog/aai-on-openrouter

wait did you mean speech to text? how do you standout?

what's the biggest advantage here?

Same here, I work for AssemblyAI and still encounter the same problem. When I call a dentist, they tell me to check with my health insurance and ask me to add us to the portal. Then I have to call the dentist again to make an appointment.

This entire cycle should be automated soon.

lol same for me, I work for AssemblyAI

and still see when I call Dentist, they tell me check with health insurance and ask them add us to portal then I call Dentist again and then make an appointment.

This whole cycle needs to be automated and will be soon!!

can't disagree, but I can say that it's worth building systems with what we have right now, and eventually whichever wins, you can just replace the API key.

Of course, that's not the same with voice, but you won't learn until you make progress.

Today we're launching the Dictation API from AssemblyAI, the first ever API built for dictation: your users speak, and it returns a response tailored to your application without any user editing required.

Here's a short clip through our Transcription API versus through the new Dictation API:

- "um so can we uh move the the meeting to thursday i think friday works better actually" - "Can we move the meeting to Friday? That works better."

Traditional transcription models are trained to be verbatim and capture everything you say.

Dictation and Voice Input models need to be aligned for a different task: a cleaned up and properly formatted input to your system or application.

- Voice typing emails and slack messages. - Doctors dictating chart notes. - Voice input to robots and hardware devices. - Coding with your voice.

These are tasks our Dictation API is built for.

We've seen a surge in demand from developers building these types of applications, and the Dictation API now makes it much easier.

It's a simple, low-cost API any developer can use to ship dictation or voice input features and apps within minutes.

- Responses in <200ms - Built on our Universal-3.5 Pro transcription model, ranked #1 for accuracy on independent benchmarks. - Transcribes 30+ languages out of the box and supports language switching. - Include custom formatting instructions on the fly - $0.62 per hour of audio, all in - no token math.

We've also open‑sourced a free Mac dictation app built on the new Dictation API, so you can experience its speed and accuracy firsthand. It's called Blurt.

I still feel on device Dication is not accurate yet.

Coming here as Developer Relations at AssemblyAI, specially where entities need to be maintained such as names from different cultures.

Can we try comparison by switching API call to our Dictation API.

By next month the competition for TTS will be even more!

Voice models are not winner take all market unlike LLM APIs

Coming here as Developer Relations at AssemblyAI

Whats the WER you saw?

I was surprised Apple talked about note taking like Granola for Apple watch but didn't integrate that or talk about for iPhone Duo.

btw. we power Granola, I work at AssemblyAI

Half of what we see or consume is hype, so we should view everything with nuance. Sometimes truth lies in the middle.