DeepgramSpeech-to-text

nova-3-general

JSON

Nova 3 General STT GA Provider default streaming, batch

01 Price

$0.26per hour

Cheaper than 76% of 236 priced speech-to-text and realtime models.

Speech-to-text and realtime $ per hour of audio, log scale 236 priced, 280 without a USD price

Speech-to-text and realtime: 236 models with a published USD price per hour of audio, from $0.05 to $20 and beyond, shown on a log scale.

other models this modelAll models by price

Plan or modePrice
pre-recorded, monolingual, Growth plan$0.22 per hour
pre-recorded, multilingual, Pay as you go$0.31 per hour
pre-recorded, multilingual, Growth plan$0.26 per hour
streaming, monolingual, Pay as you go (promo; was $0.0077)$0.29 per hour
streaming, monolingual, Growth plan (promo; was $0.0065)$0.25 per hour
streaming, multilingual, Pay as you go (promo; was $0.0092)$0.35 per hour
streaming, multilingual, Growth plan (promo; was $0.0078)$0.30 per hour

Pay-as-you-go pre-recorded monolingual list rate. Streaming rates are labelled 'Promo pricing' (no end date published). $200 free credit on Pay as you go (no expiry). Add-ons (redaction, keyterm prompting, entity detection, streaming diarization) billed extra.

Pricing source, checked 2026-11-16

02 Languages 97, 185 with regional variants

Show all 97 languages

03 Features 10 of 11 listed

Custom vocabulary
keyterm prompting (Nova-3; up to 500 tokens; mid-stream updates via Configure)
Language detection
detect_language: pre-recorded only; may fall back to next-highest model (Nova-3 > Nova-2 > Nova-1 > Enhanced > Base) if detected language unsupported
PII redaction
redact=numbers/aggressive_numbers in all pre-recorded languages; entity redaction (pci/pii/phi) English only; Nova streaming number redaction limited to 12 languages
Profanity filter
specific languages only
Punctuation
punctuate / smart_format; all available languages
Sentiment
Audio Intelligence: pre-recorded only (not streaming), English only
Speaker diarization
pre-recorded (diarizer v2) and streaming (v1 only)
Summarization
Audio Intelligence: pre-recorded only (not streaming), English only
Topic detection
Audio Intelligence: pre-recorded only (not streaming), English only
Word timestamps
every transcript returns per-word start/end times

"Not listed" means the provider's data doesn't mention it, not that the model lacks it.

04 Get this record

curl https://voicedex.dev/v1/providers/deepgram/models/nova-3-general

Free, no key needed. Use model id nova-3-general with Deepgram's own API. API reference.

Esc

↑↓ move↵ open