dylanbfox

This is very important. We do not store or copy any of the audio data you send to the API. We do not train on any audio data you send to the API. We’re a small company and absolutely believe in security and privacy.

We were caught off guard seeing ourselves on HN tonight — and our ToS is something we’re improving for when we had planned to launch in 1-2 months.

The reason why we include some of this language is because right now you can send in text data to the API for us to do transfer learning in order to create a custom language model for your use case. And soon, you can also send in audio data + transcripts to the API for us to do transfer learning on our base RNN acoustic model, to fine tune the model for your audio data.

Dylan from AssemblyAI here. We didn’t plan to launch on Hacker News for another few weeks — and are surprised to see ourselves on here — but really appreciate the interest. That’s why we don’t have more thorough benchmarks and technical specs ready yet. We totally agree that we need more information to back up our claims and are absolutely working on it. Please bear with us for now!

If you do end up trying the API now and have any feedback definitely let me know. Our goal is to work towards making the API really easy and simple for developers, but we know there’s definitely things we have to improve on.

We’ve been working hard towards launching publicly in 1-2 months but everyone on HN knows you can’t always control this stuff.

Dylan

Dylan from AssemblyAI here. Sorry for the mixup. This is just a placeholder page that we don’t link to anywhere from our site yet. I’m not sure how you found this page — but we took it down for now!

We didn’t realize our site got posted on Hacker News, and there’s probably a lot of things that are rough around the edges like this — we have been working on getting ready to do a big launch but aren’t quite there yet. Sorry for things being sloppy! Please let me know if you spot anything else: dylan[at]assemblyai.com.

This is great -- landing page and docs are super informative and makes me want to try this.

Just FYI -- getting a:

"Your connection is not secure

The owner of jitx.com has configured their website improperly. To protect your information from being stolen, Firefox has not connected to this website."

When hitting https://jitx.com/

Yup! This is in the works and should be released soon!

This is something we are looking into!

The way it works right now is that you would upload one of those old speeches or radio shows to the API, and then you'd get back a transcript that includes all the "terms of art" you customized the API to be able to recognize. Right now there's no feedback loop for you to tell the API where it was wrong, but that's something we have in mind to build! We do run QA across our entire API usage though, and from that are able to improve recognition accuracy over time. We also release updates to the models that power the API every few weeks which improves recognition accuracy too.

And then we do provide confidence scores, so we can at least give you some indication when we're not confident in the automatic transcript we're returning to you.

If you want to try out the API, you can email beta@assemblyai.com and I will look for your mail!

Good catch! Thanks!

Hey! Great question. Our system is actually able to handle transcribing "sequel" as "SQL" automatically if you were to "customize the model" for phrases like "what was my latest SQL query". It can also get words like "colonel" pronounced "kernel". In both cases, without needing the explicit pronunciation of the word. We have some customers who've uploaded thousands of proper names, for example, and we're able to transcribe all of them without needing the explicit pronunciation. This is possible because our ASR implementation is pretty different than traditional setups like Kaldi. You're right that there are some edge cases, especially with foreign words, but we're working hard on smoothing those out.

Yup! Scale has a good API for human transcription. We're an order of magnitude cheaper than human transcription services though.

We haven't actually tried that yet. I imagine if you customized your model with words from another language and then pronounced them with an english accent the API might be able to recognize them ok. Would be a fun experiment to try at least!

Thanks for taking the time to post this! On it...

Haha totally get this. We wanted to give info about how we bill (ie, per second and we don't charge for customization) but we haven't publicly disclosed our rate yet. It's pretty low (fraction of a cent per second).

We can get pretty good throughput with our setup. But the main thing about our architecture versus others is that it makes the automatic customization we offer possible. For example it takes under 2 minutes to customize the API to be able to recognize ~10,000 custom words/phrases for whatever you're building.

Thanks for the heads up on this!

Thanks for the question! If you have a stereo file, from a phone call for example, we can do separate transcripts for each channel...but we haven't launched any algorithms to auto-split speakers into the API yet! Definitely something we want to offer in the future though.

Right now we only test against a few internal test sets. On some clean speech test sets we're at ~8% WER. That's a good idea to test against some open source datasets for reporting accuracy metrics!

the interactive tools are great... very cool thanks for the write up

great write up! this is awesome

I've been curious about this too. AWS does have K80 instances available for $0.90/hour which isn't too bad for playing around and as long as they keep updating their infrastructure, you can play with the newest stuff versus having to upgrade your own all the time.

Good walk through. Have you looked into using platforms like Microsoft's LUIS, API.AI, or Wit.AI to extend the NLP?

Thanks for your feedback!

Right now this is product is just in a pilot phase, but we are absolutely looking into more international options in the future.

To your second point, what we think about a lot is an API developers could use to build this functionality into their apps or websites. So while you're in your calendar app, or in your task management app, you could easily accomplish the same thing (quickly scheduling a call and having a number provisioned). Would love your thoughts/feedback on this.

And thanks for catching that protocol mismatch!

As of now, the number that's issued is a +1 number. And it's issued for free. You have a really good idea, though. In the future we're thinking about offering the option to select the country where the call is hosted. Or to also allow people to join/call-in right from their browser. That would make it easier for international participants to join as well.

That’s a really thoughtful point. Definitely agree that the simplicity is only half of the problem.

What we think we’re seeing is a trend of more powerful networks becoming available as platforms. For developers and companies to use through things like APIs. And if that is the case, we were interested in experimenting with a simple interface to those platforms.

Swiftcall, for example, is built on top of Tropo, which is a cloud communications API that has features like multi-party conferences and add advanced features including recording and SIP.

We’re definitely keeping track of feedback about the quality of the calls and network performance, though. Because if that is poor, that’s a much more impactful and memorable bad user experience then the scheduling of the call in the first place. Agree with you there.

Thanks for taking the time to check the project out and for your thoughts!

Very cool idea. I'll look into this! Do you think it should be an add-on or come with the subscription?

Right now we're handling this on a case-by-case basis. If you have a domain you want to point to a site hosted on ClaimYourName, you can use the chat box on the bottom of the site to talk to us so we can walk you through the steps.

This is a great idea, and is something I completely overlooked. I'll be adding this in shortly!

Good idea! Agreed SSL is a good feature I should get in there quickly. Thanks for checking out the project!

Good question! I set up a microservice hosted on a static IP that talks to my Heroku app. If you're trying to do the same and want some help let me know!