Subscribe
Sign in
Deepgram
p/deepgramVoice AI platform for developers.
Start new thread
Kat Manalac

Deepgram MissionControl - Train and deploy speech AI models for free

by•

Deepgram MissionControl is the first and only platform for training end-to-end deep learning speech recognition. Train a speech model to understand your accent, product names, and more within a few clicks or API calls.

Add a comment

Replies

Best
Natalie Rutgers
Maker
Hey Product Hunt! We’ve just launched Free Accounts on Deepgram MissionControl. If you’ve ever been frustrated by inaccurate voice assistants or speech recognition APIs, MissionControl will be a pleasant surprise. We were motivated to build MissionControl after introducing Enterprise companies to our proprietary approach to speech recognition. We ditched the brittle, traditional methods to speech recognition used by legacy and Big Tech competitors for end-to-end deep learning through the entire process. That shift in approach enables us to build custom-trained speech models unique to each customer we serve. Customizing for the use case at hand means higher accuracy on the conversations that matter, even with diverse audio that includes product names, technical language, noisy environments, or speaker accents. These changes are measurable — our customers often see >20% gains in accuracy after custom training. MissionControl puts the power of custom-training directly into your hands. For the first time ever, MissionControl makes it simple to train and improve a Speech AI model without being a machine learning expert. It gives you everything you need to: - Upload and label audio data to prepare it for training. - Train a model on the audio that matters to you. - Deploy your model to get high accuracy automatic transcripts. Create your free account at missioncontrol.deepgram.com today. We know there’s a lot of opportunity for this tool, so we’d love to hear your feedback and ideas!
Liam Berryman
This is sick. Can't wait to transcribe international conference calls with this. We constantly have 10+ person calls with people coming in and out over multiple hours and it is a nightmare to try and keep up with notes. Having a custom model should increase our R&D speed significantly
Natalie Rutgers
Maker
@liam_berryman Perfect use case Liam. Most speech recognition is awful when it comes to accents. Just take a few of those recorded meetings, upload them, train a model, and you'll have significantly improved speech to text in no time.
Zachary DeWitt
Beautiful interface!
Natalie Rutgers
Maker
@zachary_dewitt Thank you Zach 😄
K ARUN
How many languages does it support ? Does it automatically detect language as well, or do we have to explicitly mention it? Can we filter out curse words in this process ??
Natalie Rutgers
Maker
@k_arun1 Our API can recognize 6 languages right now, with more becoming rapidly available. Automatically detecting languages is something that's in the works and that we're super excited about. We don't filter these out automatically right now, but it's possible to strip them out of the resulting transcript. We've dealt with labeling swear words in the past... all the combinations make for a surprisingly long list!
Dushyanth Sekhar
does this provide speaker tag for things like earning call transcripts ?