What is Deepgram

Deepgram is a voice AI platform offering Speech-to-Text, Text-to-Speech, Voice Agent, and Audio Intelligence APIs. It is aimed primarily at developers and businesses building voice-enabled products, rather than people looking for a simple recording app.

The platform supports use cases such as conversational AI, contact centers, speech analytics, and podcast transcription. Deepgram describes its services as available in real time or batch, with cloud and self-hosted deployment options. Check the official site for the deployment choices available to your particular product and plan.

For CouponCode readers, the main starting offer is a free $200 credit. This provides an opportunity to evaluate the service before committing to ongoing usage costs or a prepaid annual plan.

Key features

Deepgram covers several parts of the voice application stack. You can choose individual APIs or explore its unified Voice Agent API.

  • Speech-to-Text: Transcribe audio using models built for different workloads. Deepgram lists streaming and prerecorded processing on its pricing page.
  • Conversation-aware recognition: Flux speech recognition includes built-in turn detection and natural interruption handling for real-time voice agents.
  • Multilingual transcription: Nova models support more than 45 languages. Feature and language availability vary, so consult the developer documentation for your intended model.
  • Transcript enhancements: Listed Nova capabilities include speaker diarization, Smart Formatting, Keyterm Prompting, and Automatic Language Detection.
  • Text-to-Speech: Generate spoken output for voice applications. Deepgram presents Flux TTS as designed for live conversations, with attention to tone, expression, and continuity.
  • Unified Voice Agent API: Combine speech recognition, speech generation, and LLM orchestration in a single API instead of assembling those components separately.
  • Audio Intelligence: Deepgram also offers an API focused on insights from conversations.

When evaluating these features, test audio that resembles your actual workload. Background noise, multiple speakers, and specialized vocabulary are useful factors to include in a trial rather than relying only on a polished demo.

Plans overview

Deepgram lists three main options on its official pricing page:

  • Pay As You Go: Starts with a free $200 credit, followed by usage-based charges. The page states no minimums, no expiration, and no credit card required. It positions this option for developers and startups and includes access to all endpoints in public models.
  • Growth: Starts at $4,000 per year with prepaid credits redeemed against actual usage. Deepgram advertises savings of up to 20% and positions this plan for growing applications. It also includes access to all endpoints in public models.
  • Enterprise: Intended for businesses with large volumes, specific data or deployment requirements, or additional support needs. Pricing requires contacting sales.

Concurrency allowances differ by plan and API, so price is not the only consideration. For example, the pricing table gives Growth higher concurrency than Pay As You Go for certain streaming workloads. Check the official site for the current limits relevant to your application.

The pricing page also distinguishes rates by model and processing mode, and labels some streaming rates as promotional. Estimate costs using your expected usage and the applicable current rate rather than treating any single advertised figure as universal.

Who it's for

Deepgram is worth considering for teams that need to add speech capabilities to software they build or operate.

  • Developers and startups evaluating transcription or conversational features without immediately taking on an annual commitment.
  • Product teams integrating spoken input and generated speech into an application.
  • Contact center and analytics teams exploring transcription and conversation insights.
  • Platforms and partners embedding voice AI into their own offerings.
  • Enterprises investigating custom models, deployment requirements, or larger-scale usage.

If you only need occasional manual transcription, consider whether an API-focused platform fits your workflow. The important distinction is between using a finished consumer tool and integrating voice infrastructure into a product.

Pros & cons

Pros

  • Covers speech recognition, speech generation, and voice agents within one platform.
  • Offers a free starting credit and a usage-based entry option.
  • Provides a prepaid Growth option for teams seeking volume-related savings.
  • Lists cloud, self-hosted, and custom-model paths for differing technical needs.

Cons

  • API integration may require more technical work than a ready-made transcription interface.
  • Costs require attention to model selection, processing mode, and actual usage.
  • Growth involves an annual prepaid commitment, which may not suit uncertain demand.
  • Feature availability and concurrency limits need checking against the selected model and plan.

How to save on Deepgram

  • Deepgram $200 free credit: Start with the official free credit offer to evaluate usage before paying; check the official site for current eligibility and terms.

FAQ

Does Deepgram offer a free trial?

Deepgram advertises a free $200 credit on its pricing page. Pay As You Go then moves to usage-based charges; check the official terms before starting.

Do I need a credit card to get started?

The Pay As You Go section states that no credit card is required to start.

Which plan suits a growing application?

Deepgram positions Growth for growing applications, with annual prepaid credits and savings of up to 20%. Compare expected usage and concurrency needs before committing.

Can Deepgram power a voice agent?

Yes. Its Voice Agent API combines speech-to-text, text-to-speech, and LLM orchestration for real-time conversations.

Visit the official site: deepgram.com.