Episode Details

This story was originally published on HackerNoon at: https://hackernoon.com/how-deepgram-and-modulate-benchmark-against-real-world-audio.
We benchmarked Modulate and Deepgram on AMI, VoxPopuli, and Earnings-22. On the hardest dataset, Modulate had half the error rate at a fraction of the cost.
Check more stories related to ai at: https://hackernoon.com/c/ai. You can also check exclusive content about #speech-to-text-ai, #asr, #voice-recognition, #deepgram-vs-modulate, #deepgram-nova-3, #modulate-ai, #asr-benchmarking, #good-company, and more.

This story was written by: @modulate. Learn more about this writer by checking @modulate's about page, and for more stories, please visit hackernoon.com.

Deepgram's Nova-3 claims 5.26% WER, but on AMI it hit 28.1%. Modulate came in at 14.9% on the same dataset. We break down results across AMI, VoxPopuli, and Earnings-22, compare batch and streaming costs, and show what features each API actually includes in base pricing.