Subscribe
Sign in
Audio Transcription API by Scale — Audio transcription API powered by humans and AI
29
Replies
@rrhoover Oh woah that's really cool! Hoping more cool stuff like that gets build on top of Scale :)
@alexandrw @rrhoover I've explored a couple of these APIs to index content for https://www.findlectures.com, I'll give this one a shot to see how it compares.
One of the great things about automated transcripts is you can tell things about speaking style that aren't visible through written transcripts, e.g. how often someone says "um".
@garysieling @rrhoover That would be awesome! Find lectures looks really cool—would be really interesting to run analytics to see how lectures change over time.
🔥🔥🔥🔥 so many use cases this will be helpful for. Love the transcribing of podcasts for the hearing disabled.
@everette Really excited to see all the cool things people build with it! I would be super interested to see an asynchronous version of https://www.producthunt.com/post... for example :D
@alexandrw @everette @lucy_guo nice, saw some of the use cases on the website. Do you have some samples ? I couldnt find them on a quick glance. Wish PH supported audio/video samples on the product pages @rrhoover
@rameshdot0 @everette @lucy_guo @rrhoover Do you mean samples of our transcription? If so, then yes we do! Feel free to email me for more info alex@scaleapi.com
@whattheferguson Let's brainstorm! There have to be some really cool things we can do 🤔
Can see myself using this a lot for automating meeting note transcription, love the site too! cleanly designed :)
I've been looking for something like that to index my voice notes. Any integration with Zapier/IFTTT on the way? My notes are automatically uploaded to dropbox but I'm reluctant to write my own integration to your service.
EDIT: Wanted to add that the pricing is a bit steep for personal use. Sometimes I record a 3 minute voice note which will cost me $1.50. That's kinda high. I guess for podcasts it is reasonable. A pricing structure that includes maximum usage or maximum length would probably enable you to have more flexible plans for a variety of users.
@danr_4 Hey! Yup, we're working on ways to make it accessible in a simpler format than the API—most likely a Zapier integration.
And yup. We're actually below most of the market, but we are definitely trying to bring down the price over time as we improve our tech!
This is super cool, but hope that the 50 cents per minute rate might get more affordable in the near future ☺️
@alexandrw @trgorczynski What would the same task cost me (approximately) on Mechanical turk (obviously it would be more time-consuming, but just curious about the comparison)?
@msitver @trgorczynski The cost would be similar, but with much lower quality. To get our level of quality on MTurk, you'd likely have to create a complicated task pipeline.
Hey everyone! Really excited to be on PH once again 🎉
We've been working towards this launch and are really excited about it! We have had a lot of people ask us to build an API for audio transcription, and we're really excited about all of the potential applications 😊
Comment with a cool use case we haven't thought of and we might add it to our landing page! And if you have any questions, feel free to email me at alex@scaleapi.com.
P. S. We're also nominated for 2016 Community Product of the Year (https://www.producthunt.com/@gol...) —just a friendly reminder 🙃
@alexandrw @scaleapi @goldenkittymeow @producthunt @rrhoover hey Alexandr, how are you guys different from Trint? Also, Trint offers automated transcription as low as .17cents / minute, with your rate at .50cents / minute - how do you plan to provide a more cost effective solution?
@raj_ventures Hey Raj! We've actually tested our quality against Trint, and for now it's much higher quality (without needing some of the manual things you need to do to clean Trint transcriptions). One of the key differences is Trint is focused on building a product for one-off uses of audio transcription, whereas we want to provide a seamless API which can be easily integrated into larger-scale applications. Part of that is providing extremely accurate transcriptions which can be trusted by a machine.
That being said, one of our goals is to drive the price down as our technology becomes better, so stay tuned :)
Hey! looks like a great product, but how it actually works? how many time does it take to make a transcription ? what technologies do you use to make it more automative? @alexandrw
@mikhail_ezhov I can't give away all our secrets! But it's a simple API to send your audio files. Similar to most of our API, the default turnaround is 1 day, with options to make it 1-hour or 1-week. In terms of the technology to make it more automated, we use AI for transcription, and also build efficient tooling to make it easy for humans to verify the results.
Interesting product / features... are there any disponable API / more info on it ?
I would think lawyers, and some doctors that use dictation devices a lot, would be interested in this to transcribe depositions. Those are 2 niche areas to market to that have a lot of cash.
Product Hunt