AI Model · Captions & Transcription

MIDRank #5 in captions

Deepgram Nova-3

by Deepgram

The speed champion - real-time ASR for voice agents.

8.3/ 10.0
OUR SCORE

Price

$0.0080 / minute

Reviewed

2026-06-05T00:00:00.000000Z

Best for

Voice agents

Vendor

Deepgram

Score breakdown

quality

8/10

control

8/10

speed

10/10

value

9/10

ecosystem

8/10

Our review

Deepgram Nova-3 is the speed pick. Sub-300ms streaming ASR with strong word-error rate, optimized specifically for voice-agent and live-captioning workloads.

For pre-recorded captioning the speed advantage is invisible. For real-time UX it's transformative.

Pricing at $0.008/minute is competitive.

Verdict: best for real-time ASR. 8.3/10.

Pros

  • +Sub-300ms streaming latency
  • +Strong WER
  • +Competitive pricing

Cons

  • −Pre-recorded use cases don't benefit from speed advantage

Best for

Voice agentsLive captioningReal-time UX

Not for

Pure batch transcription (overkill)

FAQs

Other Captions & Transcription models

#1FREE

WhisperX

Community (Max Bain)

9.0/ 10.0
OUR SCORE
$0.0000 / minuteIn VideoCue
#2FREE

ElevenLabs Alignment

ElevenLabs

8.6/ 10.0
OUR SCORE
$0.0000 / minuteIn VideoCue
#3MID

OpenAI Whisper-3

OpenAI

8.4/ 10.0
OUR SCORE
$0.0060 / minuteIn VideoCue
#4MID

AssemblyAI

AssemblyAI

8.5/ 10.0
OUR SCORE
$0.01 / minute

Try Deepgram Nova-3 from the vendor.

This model isn't (yet) integrated into VideoCue - head to the official site.