Deepgram

Nova Models with 45+ Languages

Deepgram's Nova models support 45+ languages and include advanced capabilities like Speaker Diarization, Smart Formatting, Keyterm Prompting, and Automatic Language Detection. Available in streaming and pre-recorded modes.

Accurate, multilingual transcription with enhanced readability and domain-specific term recognition.

Flux TTS and Aura Voices

Flux TTS is the first voice model that thinks before it speaks, holding tone, expression, and continuity across a conversation. Aura-2 and Aura-1 offer natural, low-latency speech for voice assistants and conversational AI.

Lifelike, context-aware speech that maintains conversational flow and emotional nuance.

Unified Voice Agent API

Deepgram unifies speech-to-text, text-to-speech, and LLM orchestration into a single API, reducing complexity, latency, and cost. It natively handles interruptions, turn-taking, and conversational context.

Seamless, real-time voice agents that manage complex interactions without rigid turn-taking.

Analyze Audio for Insights

Extract actionable insights from conversational audio and text. Features include Redaction (removing PII), Entity Detection, and Speaker Diarization to enrich transcripts.

Automated compliance and deeper understanding of customer interactions.

Cloud or Self-Hosted

Deepgram is available in real time or batch, in the cloud or self-hosted, giving enterprises flexibility for data residency and compliance needs.

Deploy Voice AI in the environment that best fits your security and infrastructure requirements.

Pay As You Go and Growth

Pay As You Go requires no minimums, no expiration, and no credit card. Growth offers up to 20% savings with pre-paid annual credits. Enterprise plans are available for large volumes and custom needs.

Flexible, transparent pricing that scales with your usage.

• Unified Voice Agent APIDeepgram's Voice Agent API unifies speech-to-text, text-to-speech, and LLM orchestration into a single API, reducing complexity, latency, and cost. It supports real-time conversational AI agents that handle interruptions and take complex actions.
• Transparent, Flexible PricingDeepgram offers straightforward pricing with Pay As You Go (no minimums, no expiration, no credit card required) and Growth plans (save up to 20% with pre-paid credits). Enterprise plans are available for large volumes and custom needs.
• High-Performance Speech-to-Text ModelsNova-3 models support 45+ languages with advanced features like Speaker Diarization, Smart Formatting, Keyterm Prompting, and Automatic Language Detection. Nova-3 Monolingual is recommended for audio with background noise, crosstalk, and far-field input.
• Natural, Low-Latency Text-to-SpeechAura-2 and Aura-1 models generate natural, low-latency speech for voice assistants and conversational AI applications. Pricing is per 1k characters, with Aura-2 at $0.030/1k characters (Pay As You Go) and Aura-1 at $0.0150/1k characters.

• Streaming vs. Pre-Recorded PricingPricing for speech-to-text differs between streaming and pre-recorded audio. For example, Nova-3 Monolingual is $0.0048/min for streaming but $0.0077/min for pre-recorded (Pay As You Go). Verify which mode applies to your use case.
• Add-On CostsFeatures like Redaction, Keyterm Prompting, Entity Detection, and Speaker Diarization incur additional per-minute charges. Smart Formatting is included at no extra cost. Review the add-on pricing table to estimate total costs.
• Concurrency LimitsConcurrency limits vary by plan and API. For example, the REST API supports up to 50 concurrent requests for Speech-to-Text on Pay As You Go, while the WSS API supports up to 150. Growth plans offer higher limits (up to 225 for WSS).
• Custom Models Require ContactCustom speech-to-text models trained on proprietary datasets are available only by contacting sales. Pricing is not listed publicly and requires a custom quote.

Voice Agents

Deepgram's Voice Agent API enables real-time conversational AI agents that handle interruptions, take complex actions, and deliver natural, responsive customer interactions without delays or rigid turn-taking. It unifies speech-to-text, text-to-speech, and LLM orchestration into a single API, reducing complexity, latency, and cost.

Speech-to-Text Transcription

Deepgram's Nova models support 45+ languages with advanced capabilities including Speaker Diarization (multi-speaker detection), Smart Formatting for readability, Keyterm Prompting, and Automatic Language Detection. It powers Granola's transcription, which is fast and accurate, helping turn conversations into great meeting notes.

Text-to-Speech for Live Conversations

Flux TTS is the first voice model that thinks before it speaks—holding tone, expression, and continuity across an entire conversation. It is built for live conversations and is available for free through September 12th, with up to 45 concurrent streaming connections globally (5 in EU/AU).

Audio Intelligence

Deepgram's Audio Intelligence analyzes audio for insights, extracting actionable insights from conversational audio and text. It includes features like Redaction to automatically identify and remove sensitive PII such as social security numbers, credit cards, and phone numbers, and Entity Detection to identify and extract entities like person names, organizations, locations, and dates.

Define the exact aI Music & Audio Tools workflow Deepgram should support.

Compare it with closely related AI tools in the same category before committing.

Set review rules for accuracy, privacy, brand voice, compliance, and final approval.

Connect useful outputs to the wider stack instead of leaving them inside the AI tool.