Switch from Google STT to Deepgram for better performance at lower cost.

Deepgram measures lag in ms, not seconds, and we never round up on billing. Out of the box or a trained model, find out if Deepgram is a better fit for your Speech-to-Text (STT) needs.



All the features. Better performance. Lower cost.

Features and Capabilities
Deepgram STT
Google STT
Batch processing (1hr of audio)
30 seconds
1443 seconds
Streaming processing lag
<300 ms
800-2000 ms
General Word Error Rate (Average)
Trained Word Error Rate (Lowest)
Tailored Speech Model
Deep Search (audio)
Diarization (separate per speaker)
Up to 10
Up to 6
Audio streams processed per CPU/GPU
300 per GPU
1 per CPU
Noise Reduction
Custom Vocabulary (keyword boosting)
Profanity Filter
Numeral Formatting

Customers Love Deepgram

Deepgram customers come from a variety of innovative industries where highly accurate, scalable, real-time STT is a must-have.
Learn more

conversational AI
Conversational AI



sales enablement
Sales Enablement

contact centers
Contact Centers


No Trickery

So, when is Google the right choice?

There are times when Google is going to be the better choice for your needs. For example, if you’ve got use cases with short, command-and-response audio like, “Hey Google, tell me the weather,” or other consumer-targeted applications, it totally makes sense. Deepgram excels at business use cases that involve longer, more difficult audio rather than one person talking to a smart speaker. If that sounds like what you need, give us a try.

Contact Us

“Deepgram is doing for speech what SpaceX did for space travel. With SpaceX creating an arms race to the moon, Deepgram is creating an arms race to voice-enabled experiences.”


— Todd Fisher, CEO, CallTrackingMetrics

Talk to Us

Less talk, more research?

what is ASR
eBook: What is ASR

Download this ebook to gain a solid understanding of what Speech Recognition really is, as well as the differences between types of ASRs.

what is WER
Blog: How to calculate word error rate (WER)

Learn what Word Error Rate (WER) is, how to calculate it, and why it’s deceptive as an industry-standard metric for speech recognition.

How Deepgram Works
Whitepaper: How Deepgram Works

As more businesses embrace online channels communications, the opportunity to unlock audio data increases. Whether you are looking to unlock more value out of your call center data, build the next game-changing voice feature, or are looking to save a lot of money on speech transcription, Deepgram is the platform to get you there.