It’s great to be on Product Hunt! The Deepgram team has been hard at work bringing accurate AI-based, fast transcription to the world. And we're releasing it for free! The transcription editor in Deepgram Brain let's you: — drag & drop to upload audio and videos — we process the audio in seconds — then you can search by sound inside the file — get an automatically generated AI transcript — easily edit with synced audio to make it perfect — works best in Chrome! We hope this takes the sting out of transcribing (oh man, it's one of those things that sucks). Brain learns from the files you upload and edit, and gets better with time. There's an API and Python SDK that lets you programmatically programmatically understand your recordings. Deepgram Brain is tuned to analyze phone calls and will work on any audio you upload. We would love feedback on the user interface! It's difficult to get 'interacting with audio' to feel right. It's a next gen problem that we're tackling head on. There's keyboard shortcuts and two types of cursors: a Text Cursor and an Audio Cursor (the Deepgram Dot, kinda like karaoke). Transcribing is also a great way to learn new things. Upload a lecture and transcribe it. Then post add it to Youtube subtitles for the lectures. Others will thanks you and you'll have have better recall later :). Glad to answer questions about speech and brain!

I got a 404 on the main PH link (brain.deepgram.com), and then if I try signing up through deepgram.com I get a never ending circle... Actually, worked out it was because I put "Don't call me, I live in Australia and you'd probably just dial at weird hours" in the phone number field... validation bug? But yeah, I'd say phone number field would deter people from signing up. LOVE the idea of this though. 3 months ago there was almost none of these, and now there's been like 3-4 posted, but most of the cost per word on them has been a bit high for my liking (particularly given they were using free APIs). * Uploading files UI: I almost didn't see the bar at the bottom * I assume servers can transcribe faster than I can listen, so I like that this does it all on the back-end without me having to listen to an hour of audio (some of them make me do this) * That said, in the files page, If something is processing, it'd be nice to see a bar to indicate how much of the processing was done, and an expected time of completion - even if it's just a guestimate. * AI thing: Can you give me more info on your process? A couple of the other transcription services combine multiple Transcription AI's (Watson, Google Voice, etc.) to create a single "best guess" transcription. What's your process? * Transcribing thing: Can you pick up multiple speakers? I saw the IBM Watson demo will do that (and that's really useful for me) Okay, so while typing this, the 1hr / 15Mb MP3 conversation I uploaded has transcribed........ and I honestly don't know what to say. About 2/3 of the conversation is missing from the transcription. It's not studio quality audio, but I record on a Blue Yeti Pro and the other speaker was fairly understandable... even if it couldn't understand either of us in parts, I'd expect wrong words not big gaps.

Looking forward to testing your platform for audio and video transcriptions and accessing captions files - in the meantime I have added you to my curations.

Do you plan on exposing alignment data via the API?

I explored a few of these services to get transcripts for https://www.findlectures.com, and Deepgram is by far the most pleasant REST API to work with - glad to see they are expanding their offerings!

All is ok.

Pros:

It's free. And it transcribes good.

Cons:

I haven't noticed yet