Deepgram
Transform speech to text or voice effortlessly, in 36 languages.. [Freemium]
Last verified:
What is Deepgram?
Deepgram is an enterprise Voice AI platform that provides real-time APIs for speech-to-text (STT), text-to-speech (TTS), voice agents, and audio intelligence. As a foundational AI company, Deepgram focuses on building voice technology that transforms human-to-machine interactions, powering every conversation through advanced, contextualized voice AI models built for the real world.
Key features include the Flux Multilingual conversational STT model supporting 10 languages (English, Spanish, German, French, Hindi, Russian, Portuguese, Japanese, Italian, Dutch), Nova-3 models for monolingual and multilingual transcription with 45+ language support, speaker diarization for multi-speaker detection, PII redaction, keyterm prompting for domain-specific jargon, smart formatting for punctuation and casing, and a unified Voice Agent API that combines STT, TTS, and LLM orchestration. Audio Intelligence features include summarization, topic detection, sentiment analysis, and intent recognition. The platform offers both cloud API and self-hosted/on-premise deployment options.
Deepgram is designed for developers, product teams, startups, and enterprises building voice applications such as call centers, meeting software, voice assistants, conversational AI applications, healthcare patient engagement systems, QSR drive-thru automation, and any application requiring real-time transcription or voice interactions. The platform is trusted by both startups and enterprises worldwide.
Deepgram pricing
Pricing model: Freemium
Deepgram offers a Pay-As-You-Go plan with $200 free credit and no credit card required, plus a Growth plan with pre-paid yearly credits starting at $4K+/year that saves up to 20%. Speech-to-Text pricing: Flux English at $0.0065/min (Pay As You Go) or $0.0057/min (Growth), Flux Multilingual at $0.0078/min or $0.0068/min, Nova-3 Monolingual at $0.0048/min or $0.0042/min, Nova-3 Multilingual at $0.0058/min or $0.0050/min. TTS pricing: Aura-2 at $0.030/1k characters or $0.027/1k characters, Aura-1 at $0.0150/1k characters or $0.0135/1k characters. Voice Agent API: Standard at $0.075/min or $0.068/min, Advanced at $0.163/min or $0.146/min, with BYO TTS and BYO LLM options available. Audio Intelligence (Summarization, Topic Detection, Sentiment Analysis, Intent Recognition) at $0.0003/1k input tokens - $0.0006/1k output tokens or $0.00024/$0.00048 for Growth. Add-ons include Redaction at $0.0020/min or $0.0017/min, Keyterm Prompting at $0.0013/min or $0.0012/min, Speaker Diarization at $0.0020/min or $0.0017/min, and Smart Formatting included. Enterprise plans available through Contact Sales for large volumes or custom deployment needs.
Deepgram pros
- Excellent speech-to-text accuracy with Nova-3 and Flux models
- Strong developer experience with official SDKs for Python, Node.js, Go, .NET, Java, C#
- Good real-time streaming performance with low latency
- Enterprise-grade reliability and SOC 2 Type 1 & Type 2 certified
- Competitive pay-as-you-go pricing with $200 free credit
- On-premise and self-hosted deployment options for data control
- Unified Voice Agent API reduces complexity, latency, and cost
- Speaker diarization detects and labels multiple speakers
- PII redaction automatically removes sensitive data like credit cards and SSNs
- Keyterm prompting boosts accuracy for domain-specific jargon up to 90%
- Smart formatting automatically handles punctuation, casing, dates, currency
- HPAA compliant with Business Associate Agreements for enterprise customers
- GDPR ready with dedicated EU data residency endpoint
- Flux Multilingual supports conversational STT in 10 languages without routing
- Audio Intelligence includes summarization, topic detection, sentiment analysis, intent recognition
- No credit card required to start with free tier
- Supports 45+ languages for speech-to-text transcription
- Up to 225 concurrent streams for WSS API on Growth plan
- Voice Agent API handles interruptions and complex actions naturally
Deepgram cons
- Fewer voice options than dedicated TTS platforms
- Better known for transcription than text-to-speech synthesis
- Does not support emotion instructions for TTS
- Limited languages available compared to some competitors
- Does not offer voice cloning capabilities
- Some languages lack full feature support across all models
- Keyterm Prompting limited to 500 tokens per request
- Maximum file size of 2 GB for pre-recorded audio
- Rate limits of 100 concurrent requests per project for Nova/Base/Enhanced models
- Requests exceeding 10 minutes return Gateway Timeout error for Nova/Base/Enhanced
Frequently asked questions about Deepgram
How much does Deepgram Speech-to-Text cost per hour?
Deepgram charges per minute of audio, not per hour. For example, Nova-3 Monolingual costs $0.0048/min on Pay As You Go, which equals $2.88/hour. Flux English costs $0.0065/min ($3.90/hour), and Flux Multilingual costs $0.0078/min ($4.68/hour). The Growth plan offers discounted rates.
What is included in the $200 free credit?
The $200 free credit on the Pay-As-You-Go plan can be redeemed against actual usage for any Deepgram endpoint including Speech-to-Text, Text-to-Speech, Voice Agent API, and Audio Intelligence. No credit card is required to sign up, and credits do not expire.
What is the difference between Pay-As-You-Go and Growth plans?
Pay-As-You-Go has no minimums, no expiration, and no credit card required with $200 free credit. Growth plan requires pre-paid credits for the year ($4K+), saves up to 20% on usage, offers higher concurrency limits (up to 225 for WSS API vs 150), and both provide access to all endpoints in public models.
Does Deepgram charge for silence or round up audio time?
Deepgram charges based on actual audio processing time. The pricing is per minute of audio processed, and costs are calculated based on the actual duration of speech processed through their APIs.
How does pricing work for the Voice Agent API?
Voice Agent API pricing is per minute of conversation. Standard tier is $0.075/min (Pay As You Go) or $0.068/min (Growth). Standard with BYO TTS is $0.065/min or $0.051/min. Custom with BYO LLM is $0.056/min or $0.059/min. Custom with BYO LLM + TTS is $0.050/min or $0.041/min. Advanced tier is $0.163/min or $0.146/min, with Advanced - BYO TTS at $0.122/min or $0.110/min.
Are there extra fees for real-time streaming vs. pre-recorded audio?
No extra fees differentiate real-time streaming from pre-recorded audio. The same per-minute pricing applies to both REST API (batch/pre-recorded) and WSS API (real-time streaming). Concurrency limits differ: up to 50 concurrent REST API requests vs up to 150-225 WSS API requests depending on plan.
Do Audio Intelligence features cost extra?
Yes, Audio Intelligence features (Summarization, Topic Detection, Sentiment Analysis, Intent Recognition) are billed separately based on token usage at $0.0003/1k input tokens - $0.0006/1k output tokens on Pay As You Go, or $0.00024/$0.00048 on Growth plan. These are add-ons to base transcription costs.
Is Text-to-Speech (Aura) billed by the minute or character?
Text-to-Speech using Aura models is billed by character, not by minute. Aura-2 costs $0.030/1k characters on Pay As You Go or $0.027/1k characters on Growth. Aura-1 costs $0.0150/1k characters on Pay As You Go or $0.0135/1k characters on Growth.
Can I deploy Deepgram on-premise or in a private cloud?
Yes, Deepgram offers self-hosted/on-premise deployment options for organizations requiring data residency guarantees, enhanced security, or compliance requirements. The self-hosted offering includes the same API and feature set as the hosted API with no feature limits, available for speech-to-text, text-to-speech, and language understanding.
Is Deepgram HIPAA and SOC 2 compliant?
Yes, Deepgram is SOC 2 Type 1 & Type 2 certified with independently audited security controls. For HIPAA compliance, Deepgram signs Business Associate Agreements (BAA) for Enterprise customers handling electronic Protected Health Information (ePHI). The platform is also GDPR Ready with EU data residency, CCPA compliant, and PCI compliant.