deepgram.com

Flux is Deepgram's conversation-aware speech platform: Flux STT is speech-to-text built to listen for turn-taking and interruptions in live dialogue, and Flux TTS is text-to-speech built to hold tone, expression, and continuity across a conversation. Both are positioned for voice agents and are available in real time or batch, in the cloud or self-hosted. They differ from standard STT/TTS in that they are designed around the back-and-forth of a conversation rather than one-shot transcription or narration.

Flux STT: listening for the conversation, not just the words

Standard speech-to-text turns audio into a transcript. Flux STT is described as conversational speech recognition that detects end-of-turn and interruptions and streams transcripts in real time. That matters for a voice agent because the agent needs to know when the user has finished speaking before it responds, and it needs to cope with being talked over.

The practical effect: instead of building your own silence detection and barge-in logic on top of a generic transcript stream, you get signals designed for turn-taking from the recognition layer itself.

Flux TTS: speaking with continuity

Flux TTS is described as the first voice model that "thinks before it speaks," holding tone, expression, and continuity across an entire conversation. Standard TTS typically synthesizes each response independently, so tone can drift between turns. Flux TTS is aimed at keeping the voice consistent as the dialogue progresses.

Deepgram lists latency as low as 80ms for its text-to-speech offering, and positions Flux TTS for live conversations rather than pre-recorded narration.

Where Flux fits in the voice agent stack

Deepgram describes a single, unified Voice Agent API that combines speech-to-text, text-to-speech, and LLM orchestration, rather than stitching separate components together. The flow is:

1. Audio input — a partner-provided transport layer (including telephony partners via PSTN/SIP) bridges streamed audio into Deepgram's APIs and plays TTS audio back to the user.
2. Flux STT — detects end-of-turn and interruptions and streams transcripts in real time.
3. LLM orchestration — manages context and memory, system and prompt updates, response injections, and function calling, connecting to language models via API.
4. Flux TTS — converts the AI response into natural-sounding speech.
5. Business logic and external systems — your application logic executes domain-specific actions and integrates external APIs, with results fed back into the conversation context.

Availability and access

• Modes: real time or batch.
• Deployment: cloud or self-hosted.
• Flux TTS free period: build with Flux TTS free through September 12th, with up to 45 concurrent streaming connections in the US and 5 in the EU and AU.
• Promotion: spend $1 on Flux TTS, get $1 back, up to $500.

Deepgram also offers a free trial and demo requests. Pricing beyond the promotion is not specified in the available material, so confirm current terms directly with Deepgram before committing to a production plan.

How to decide whether Flux is the right fit

A concrete way to test the difference: build a short agent that lets a user interrupt mid-sentence. With standard STT you would typically add your own voice-activity detection and barge-in rules; with Flux STT, end-of-turn and interruption detection are part of the recognition layer, so you can check whether the agent stops speaking and responds appropriately without extra plumbing.

Deepgram is an enterprise voice AI platform that provides speech-to-text (STT), text-to-speech (TTS), a unified Voice Agent API, and Audio Intelligence. It is aimed at teams building real-time voice agents and other production voice applications that need accurate transcription, low-latency speech output, and a way to run the full conversation loop through one API. The platform is available in real time or batch, in the cloud or self-hosted.

Core products

Deepgram describes its platform as “one platform to power your entire voice AI stack,” organized around four capabilities:

The newest generation of models is branded Flux. Deepgram describes Flux as “the first conversation-aware speech platform,” built to natively understand turn-taking, handle interruptions, and carry conversational context. Flux STT listens; Flux TTS speaks.

Flux TTS

Flux TTS is positioned as text-to-speech built for live conversations — a voice model that “thinks before it speaks,” holding tone, expression, and continuity across an entire conversation. Deepgram lists a promotional offer on its site: spend $1 on Flux TTS and get $1 back, up to $500. It also states that Flux TTS can be built with free through September 12th, with up to 45 concurrent streaming connections in the US and 5 in the EU and AU. Those limits and dates are time- and region-specific, so confirm current terms before planning capacity around them.

The unified Voice Agent API

Rather than stitching together separate components, Deepgram unifies speech-to-text, text-to-speech, and LLM orchestration into a single API, which the company says reduces complexity, latency, and cost. The conversation flow works like this:

1. User — a customer or end user speaks naturally.
2. Audio input — a partner-provided transport layer bridges streamed audio into Deepgram’s APIs and plays back TTS audio to the user, including telephony partners via PSTN/SIP.
3. Speech-to-text — conversational speech recognition detects end-of-turn and interruptions and streams transcripts in real time.
4. LLM orchestration — coordinates context, memory, and AI reasoning. It maintains conversation history and system settings, adjusts role or behavior, refines instructions in real time, injects guidance or context, calls functions to execute actions or fetch data, and connects to language models via API for response generation.
5. Text-to-speech — converts AI responses into natural-sounding speech.
6. Business logic — your custom application logic executes domain-specific actions, integrates external APIs, or adds validation before results flow back into the agent.
7. External systems — databases, APIs, and back-office systems the agent interacts with; results are fed back into the conversation context for grounding.

The practical takeaway: if you are building a voice agent, the orchestration layer that normally sits between STT, an LLM, and TTS is handled inside the API, and your application supplies transport, business logic, and system integrations.

Deployment and fit

Deepgram states the platform is available in real time or batch, in the cloud or self-hosted. That matters if you have data-residency or infrastructure requirements that rule out a fully hosted setup, or if you need batch processing for recorded audio alongside live conversations.

The platform is marketed to enterprises — the site says it is “trusted by the world’s top companies” and emphasizes solutions that scale. If your workload is a single prototype, the free trial and demo paths are the natural entry points; if you are productionizing a contact-center or telephony agent, the PSTN/SIP transport support and concurrency limits are the details to verify against your expected call volume.

How to evaluate it for your use case

• If you need a full voice agent stack: start with the Voice Agent API, since it bundles STT, TTS, and LLM orchestration and removes the work of integrating three vendors.
• If you only need transcription: Speech-to-Text covers real-time and batch use.
• If voice quality and latency drive your product: compare Flux TTS against your current provider, paying attention to the stated latency floor (as low as 80ms) and the concurrency caps during any promotional period.
• If you need conversation analytics: Audio Intelligence extracts insight from conversations without a separate pipeline.

Deepgram offers a free trial and a demo request on its site; use those to test accuracy on your own audio and to confirm current pricing, concurrency, and regional availability before committing.

Deepgram offers four product paths—Speech-to-Text (STT), Text-to-Speech (TTS), Voice Agent, and Audio Intelligence—and the right choice depends on what your application needs to do with audio. If you need to transcribe speech, start with STT. If you need to generate speech, start with TTS. If you're building a full conversational agent that listens, thinks, and responds, the unified Voice Agent API bundles STT, TTS, and LLM orchestration into one call. If you need insight from conversations rather than live interaction, Audio Intelligence is the fit. The sections below break down each path, deployment options, and real-time versus batch considerations so you can match a product to your use case.

Start With the Job You Need Done

The fastest way to narrow the choice is to identify the primary task your application performs.

These products can also be combined. The Voice Agent API is essentially a packaged combination of STT, TTS, and orchestration, so if you're building an agent, you generally don't need to stitch the pieces together yourself.

Speech-to-Text: When Transcription Is the Core Need

Choose STT when your product's value comes from turning audio into text—for example, call transcription, captioning, meeting notes, or voice search. Deepgram describes its STT as conversational speech recognition that detects end-of-turn and interruptions and streams transcripts in real time.

This matters most for voice agents and live conversations, where knowing when a speaker has finished talking is as important as the words themselves. If your use case involves turn-taking or interruptions, STT with conversational awareness is the relevant capability to evaluate.

Text-to-Speech: When Voice Output Is the Core Need

Choose TTS when you need to generate spoken output—for example, reading responses aloud, voice notifications, or giving an AI assistant a voice. Deepgram positions its Flux TTS as text-to-speech built for live conversations, holding tone, expression, and continuity across an entire conversation.

The stated latency floor is as low as 80ms, which is the figure to weigh if your application is latency-sensitive. For non-conversational output, such as pre-generated audio or batch voice rendering, the conversational features of Flux TTS may be less critical than voice quality and cost.

Voice Agent API: When You're Building a Conversational Agent

Choose the Voice Agent API when you want a complete listen-think-speak loop without assembling separate components. Deepgram states that it unifies speech-to-text, text-to-speech, and LLM orchestration into a single API, reducing complexity, latency, and cost.

The documented flow is:

1. User speaks — your customer or end user speaks naturally.
2. Audio input — a partner-provided transport layer bridges streamed audio into Deepgram's APIs and plays back TTS audio to the user. This includes telephony partners via PSTN/SIP.
3. Speech-to-text — conversational recognition detects end-of-turn and interruptions and streams transcripts in real time.
4. LLM orchestration — coordinates context, memory, and AI reasoning, including conversation history, system updates, prompt updates, response injections, function calling, and connection to language models via API.
5. Text-to-speech — converts AI responses into natural-sounding speech.
6. Business logic — your custom application logic executes domain-specific actions, integrates external APIs, or adds validation before results flow back into the agent.
7. External systems — databases, APIs, and back-office systems the agent interacts with, with results fed back into the conversation context for grounding.

The practical implication: if your team would otherwise spend engineering effort integrating a separate STT vendor, TTS vendor, and LLM orchestration layer, the unified API is the path that removes that integration work. If you already have a working pipeline and only want to swap one component, a standalone STT or TTS product may be simpler.

Audio Intelligence: When You Need Insight, Not Interaction

Choose Audio Intelligence when the goal is understanding conversations rather than participating in them—for example, analyzing support calls, extracting topics, or summarizing recorded interactions. Deepgram describes it as insight from every conversation. This is a different job from live agents: it sits on top of transcription and adds analysis rather than generating responses.

Deployment: Cloud or Self-Hosted

Deepgram states its platform is available in real time or batch, in the cloud or self-hosted. The choice depends on your constraints:

• Cloud is the default path when you want managed infrastructure and faster setup.
• Self-hosted is relevant when you have data residency, network, or compliance requirements that keep audio on your own infrastructure.

The source material does not specify which products are available under which deployment model, so confirm self-hosting support for your specific product before committing.

Real-Time vs. Batch

Match this to your latency requirement:

• Real-time is required for live conversations, voice agents, and any interaction where the user is waiting for a response. STT streaming and low-latency TTS are the relevant capabilities here.
• Batch fits recorded audio, post-call analysis, and workloads where turnaround time is measured in minutes rather than milliseconds.

A single application can use both—for example, real-time STT during a call and batch processing afterward for analysis.

A Practical Selection Sequence

1. Define the primary task — transcribe, speak, converse, or analyze.
2. Check the latency requirement — if a human is waiting on a response, prioritize real-time capabilities; otherwise batch is sufficient.
3. Decide build vs. assemble — if you want one API for the full conversation loop, evaluate Voice Agent; if you only need one component, evaluate STT or TTS alone.
4. Confirm deployment — verify cloud or self-hosted availability for your chosen product against your data and infrastructure constraints.
5. Test against your own audio — Deepgram offers a free trial and demo request; validate accuracy and latency with representative samples before committing.

Common Pitfalls to Avoid

• Choosing Voice Agent when you only need transcription. The unified API adds orchestration you may not need; standalone STT is the leaner option.
• Ignoring turn-taking. For conversational use cases, end-of-turn detection and interruption handling are core requirements, not extras—confirm they're covered.
• Assuming deployment flexibility applies to every product. The source confirms cloud and self-hosted availability at the platform level but does not map it to each product; verify before you plan around it.
• Skipping a trial with your own data. Accuracy claims are production-oriented, but your audio conditions (telephony, accents, background noise) are what determine real performance.

Deepgram's Voice Agent API combines speech-to-text, text-to-speech, and LLM orchestration in a single API call instead of requiring you to stitch together separate components. The flow runs from the user's audio through STT, LLM orchestration, and TTS, with your business logic and external systems able to feed context back into the conversation. This design is intended to reduce complexity, latency, and cost for teams building real-time voice agents.

The end-to-end flow

The platform describes the pipeline as a sequence of connected stages:

What LLM orchestration handles

The orchestration layer is where conversation state and control live. According to the platform description, it manages:

• Context: maintains conversation history and system settings.
• System updates: adjusts role or behavior.
• Prompt updates: refines instructions in real time.
• Response injections: adds guidance or context.
• Function calling: executes actions or fetches data.
• LLM connection: connects to language models via API for response generation.

Why the unified approach matters

Stitching separate STT, LLM, and TTS services together means managing multiple integrations, passing state between them, and absorbing the latency of each handoff. Deepgram's stated benefit of the single API is reduced complexity, latency, and cost. The STT layer is built for conversation specifically — it detects end-of-turn and interruptions — which is the behavior a voice agent needs to respond naturally rather than waiting for a fixed silence threshold.

Where Flux STT and Flux TTS fit

The platform's current models are Flux STT and Flux TTS, described as the first conversation-aware speech platform built to natively understand turn-taking, handle interruptions, and carry conversational context. They are available in real time or batch, in the cloud or self-hosted.

• Flux STT listens; Flux TTS speaks.
• Flux TTS is described as holding tone, expression, and continuity across an entire conversation, with latency as low as 80ms for text-to-speech on the platform.
• A promotional offer states that spending $1 on Flux TTS returns $1 back, up to $500, and that Flux TTS can be built with free through September 12th, with up to 45 concurrent streaming connections in the US and 5 in the EU and AU.

Choosing your path

The platform frames adoption as choosing the Voice AI journey that fits your business. In practice, the decision comes down to scope:

• Use the unified Voice Agent API if you want one integration covering STT, orchestration, and TTS, and you want the platform to manage turn-taking and context.
• Use individual APIs (speech-to-text, text-to-speech, audio intelligence) if you already have orchestration in place or only need one capability.

For teams evaluating the stack, the practical next step is to test the conversational behavior directly — particularly interruption handling and end-of-turn detection — since those determine whether a voice agent feels natural in production.

You can try Deepgram at no cost through two routes shown on its site: a general "Try for free" option and a "Get a demo" request. If you specifically want to build with the new Flux TTS model, the site states it is free to build with through September 12th, with a cap of up to 45 concurrent streaming connections in the US and 5 in the EU and AU. Those are the only free-access details the site provides; pricing for ongoing use beyond the trial or the Flux TTS window is not listed, so treat continued access as something to confirm with Deepgram.

What you can try

Deepgram presents one platform covering the full voice AI stack:

The site also mentions a Flux TTS promotion: spend $1 on Flux TTS and get $1 back, up to $500.

Start with the free trial or demo

1. Go to deepgram.com and choose Try for free to begin building, or Get a demo if you want a guided walkthrough first.
2. Pick the product that matches your task — for example, STT if you need transcription, or the Voice Agent API if you want an end-to-end conversational agent.
3. If you are testing Flux TTS during the free window, keep your concurrent streaming connections within the stated limits (45 in the US, 5 in the EU and AU) so you stay inside the free build terms.
4. Verify your setup by running a short real-time session and checking that transcripts stream and TTS audio plays back as expected.

How the unified Voice Agent API fits in

Rather than stitching together separate components, Deepgram combines STT, TTS, and LLM orchestration into one API. In the flow the site describes:

• Audio input comes from a transport layer (including telephony partners via PSTN/SIP) that bridges streamed audio into Deepgram's APIs and plays TTS audio back to the user.
• STT recognizes conversational speech, detects end-of-turn and interruptions, and streams transcripts in real time.
• LLM orchestration manages context, memory, and reasoning — conversation history, system and prompt updates, response injections, and function calling — connecting to language models via API.
• TTS converts responses into natural-sounding speech.
• Business logic and external systems execute domain-specific actions and feed results back into the conversation context.

This matters for a trial because a single API call replaces several integrations, which reduces complexity, latency, and cost — and gives you fewer moving parts to validate before deciding whether to continue.

Common sticking points

• Region limits on Flux TTS: the free build window caps concurrent streaming connections by region, so a US-based test can scale further than one in the EU or AU.
• Time-bound offer: the Flux TTS free build period ends September 12th; plan your evaluation before then if that model is your focus.
• Deployment choice: the platform is available in real time or batch, in the cloud or self-hosted — confirm which option your trial uses before benchmarking latency.
• Beyond the trial: the site does not publish ongoing pricing, so ask about costs for production scale rather than assuming the free terms carry over.