Deepgram

Deepgram is a speech AI platform with fast, accurate transcription APIs for developers. Its Nova-3 model transcribes pre-recorded audio at $0.0043 a minute and streaming audio at $0.0077 a minute, with support for more than 36 languages. It is known for low latency, which makes it popular for voice agents and live captioning.

Key Features

Fast speech-to-text API with Nova-3
Real-time streaming transcription
$200 free credit, no card required
36+ languages supported
Text-to-speech and voice agents too
Low latency for live applications

Pricing

Plan	Price	Details
Free Credit	$0	$200 in free credits, no card required; covers over 700 hours of Nova-3 transcription.
Nova-3 Batch	$0.0043/min	Pre-recorded transcription with the Nova-3 model, billed per minute.
Nova-3 Streaming	$0.0077/min	Real-time streaming transcription with the Nova-3 model, billed per minute.

Prices as listed in 2026 research. Always check the official site for the latest price — plans change. Free plan: Yes — it has a free plan.

Best for: developers who need fast, low-latency transcription for real-time and batch use.
Popular with: Developers, Business Owners

How it works

A $200 free credit with no card required lets you transcribe tens of thousands of minutes before paying, which is one of the most generous free offers among speech APIs. Deepgram also offers text-to-speech and voice agent tools, so it can cover both sides of a voice product.

Deepgram is a top choice for developers building voice agents, live transcription, or batch processing at scale. The API is fast and well documented, and the free credit is enough to build and test a real product. For a one-off transcript, a consumer tool is simpler, but for production speech AI, Deepgram is hard to beat.

Key Features

Pricing

How it works

Pros and Cons

✓ Pros

✗ Cons

Related Tools

Otter.ai

Fireflies.ai

Fathom