Voice Search Partnership with Algolia

AssemblyAI Speech-to-Text API update includes the launch of Voice Search with Algolia, improved accuracy, and enhanced speed.

Voice Search Partnership with Algolia

This month we had one of our most anticipated launches with our friends, and the leaders in online search, Algolia. In addition, we had another model update that has shown significant improvements in accuracy and turnaround time (speed) for phone calls, videos, and podcasts. Finally, check out our two of our newest features focused around security: PII Audio Redaction and Profanity Filtering.

Voice Search with Algolia

Our friends at Algolia are known around the globe for their super fast and robust search that powers popular websites like Twitch, Stripe, and Lacoste. As a leader in search, Algolia understands the importance of giving Voice Search capabilities to their customers.

Since 71% of users prefer to search through voice compared to using a keyboard, we wanted to make it easy add voice search to your website or app in seconds. It’s in that spirit that we’re happy to announce that we’ve teamed up with Algolia to launch a plugin for their popular InstantSearch.js library. With this library, developers can add accurate Voice Search to their websites in just a few lines of code.

See the full API documentation here.

Further Accuracy Improvements

With our most recent update, we are consistently out-benchmarking all other speech-to-text API's in accuracy on our asynchronous English model. Below are our average accuracy %'s (using Word Error Rate or WER) based on benchmark reports run this month:

Speed Enhancements

Speech-to-text turnaround time was a big focus for us this month to improve speed for customers looking to transcribe webinars, video interviews, and conference calls - all longer forms of audio and video. You should notice a ~22% increase in speed across all of our models due to a couple of updates we made:

  • Optimizations in our model size
  • Improvements in our beam search

PII Redaction for Audio

In addition to PII redaction on transcripts, we now send back an updated audio URL that includes the same redactions being made in the audio clip. This means that your audio file will now silence any personal identifiable information (PII) like social security numbers and credit cards from playback using the new URL we send you.

Profanity Filter

Looking to remove curse words from your transcription text? Our new profanity filter allows you to quickly remove curse words and replace them with "#" characters in the transcript so you don't have to worry about it!