STT

Speech-to-text providers supported by Patter.

STT (Speech-to-Text)

STT is used in pipeline mode to transcribe caller audio before it reaches your LLM. If you use an engine such as OpenAIRealtime or ElevenLabsConvAI, speech recognition is handled internally by the engine.

Every STT class is imported by name from the package barrel: import { DeepgramSTT } from "getpatter".

Quickstart

// npx tsx example.ts
import { Patter, Twilio, DeepgramSTT, ElevenLabsTTS } from "getpatter";

const phone = new Patter({ carrier: new Twilio(), phoneNumber: "[phone]" });  // TWILIO_* from env

const agent = phone.agent({
  stt: new DeepgramSTT({ endpointingMs: 80 }),      // DEEPGRAM_API_KEY from env
  tts: new ElevenLabsTTS({ voiceId: "rachel" }),    // ELEVENLABS_API_KEY from env
  systemPrompt: "You are a helpful assistant.",
  firstMessage: "Hi!",
});

await phone.serve({ agent });

Supported providers

Flat import	Namespaced import	Env var
DeepgramSTT	getpatter/stt/deepgram	DEEPGRAM_API_KEY
WhisperSTT	getpatter/stt/whisper	OPENAI_API_KEY
OpenAITranscribeSTT	getpatter/stt/openai-transcribe	OPENAI_API_KEY
CartesiaSTT	getpatter/stt/cartesia	CARTESIA_API_KEY
AssemblyAISTT	getpatter/stt/assemblyai	ASSEMBLYAI_API_KEY
SonioxSTT	getpatter/stt/soniox	SONIOX_API_KEY
SpeechmaticsSTT	getpatter/stt/speechmatics	SPEECHMATICS_API_KEY

SpeechmaticsSTT is being ported to TypeScript in the upcoming release — see the ## Unreleased section in CHANGELOG.md. Use Python or wait for the next minor version.

Speechmatics is supported by the Python SDK but not yet by the TypeScript SDK — use the Python SDK if you need Speechmatics.

Model enums

Each provider exports a typed const-object of valid model IDs alongside the provider class. They keep model options tab-completable and reject typos at construction time, while still accepting raw strings for forward compatibility:

import { DeepgramSTT, DeepgramModel } from "getpatter";

const stt = new DeepgramSTT({ model: DeepgramModel.NOVA_3 });

The same pattern applies to AssemblyAIModel, CartesiaSTTModel, and SonioxModel.

Deepgram

Streaming STT backed by Deepgram’s nova-3 model.

import { DeepgramSTT } from "getpatter";

const stt = new DeepgramSTT();                                    // reads DEEPGRAM_API_KEY
const stt = new DeepgramSTT({ apiKey: "dg_...", endpointingMs: 80 });

Parameter	Type	Default	Description
apiKey	string	—	API key — reads from DEEPGRAM_API_KEY if omitted.
language	string	"en"	BCP-47 language code.
model	string	"nova-3"	Deepgram model ID.
encoding	string	"linear16"	Audio encoding sent to Deepgram.
sampleRate	number	16000	Sample rate in Hz.
endpointingMs	number	150	Utterance endpointing in milliseconds.
utteranceEndMs	number | null	1000	Grace period after speech ends.
smartFormat	boolean	false	Smart formatting (numbers, dates, punctuation). Defaults to false because telephony agents feed transcripts straight back into an LLM, where smart-format rewrites can confuse downstream tool-call argument parsing. Pass smartFormat: true to opt back in.
interimResults	boolean	true	Stream interim transcripts.
vadEvents	boolean	true	Emit VAD start/end markers.

Whisper (OpenAI)

HTTP-based STT via OpenAI Whisper. Reuses OPENAI_API_KEY.

import { WhisperSTT } from "getpatter";

const stt = new WhisperSTT();                                     // reads OPENAI_API_KEY
const stt = new WhisperSTT({ apiKey: "sk-...", language: "es" });

Whisper on mulaw 8 kHz routinely hallucinates short fillers ("you", ".", "thank you"). For production prefer OpenAITranscribeSTT (gpt-4o-transcribe) — same OPENAI_API_KEY, ~10× faster, no hallucination floor.

OpenAI Transcribe (gpt-4o-transcribe)

First-class STT for OpenAI’s gpt-4o-transcribe and gpt-4o-mini-transcribe models — drop-in replacement for WhisperSTT with stronger multilingual quality and significantly lower latency. Reuses OPENAI_API_KEY.

import { OpenAITranscribeSTT } from "getpatter";

const stt = new OpenAITranscribeSTT();                                  // reads OPENAI_API_KEY, defaults to gpt-4o-transcribe
const stt2 = new OpenAITranscribeSTT({ model: "gpt-4o-mini-transcribe" }); // cheaper variant
const stt3 = new OpenAITranscribeSTT({ apiKey: "sk-...", language: "es" });

Parameter	Type	Default	Description
apiKey	string	—	API key — reads from OPENAI_API_KEY if omitted.
language	string	—	BCP-47 language code. Auto-detect when omitted.
model	string	"gpt-4o-transcribe"	Either "gpt-4o-transcribe" or "gpt-4o-mini-transcribe".
responseFormat	string	"json"	Pass "verbose_json" to expose segment-level confidence and timestamps.

Cartesia

Streaming STT using Cartesia’s ink-whisper. See Cartesia setup.

import { CartesiaSTT } from "getpatter";

const stt = new CartesiaSTT();                                    // reads CARTESIA_API_KEY
const stt = new CartesiaSTT({ apiKey: "csk_...", language: "en" });

AssemblyAI

Universal Streaming STT via the AssemblyAI v3 WebSocket API. See AssemblyAI setup.

import { AssemblyAISTT } from "getpatter";

const stt = new AssemblyAISTT();                                  // reads ASSEMBLYAI_API_KEY

Soniox

Real-time STT via Soniox.

import { SonioxSTT } from "getpatter";

const stt = new SonioxSTT();                                      // reads SONIOX_API_KEY

Missing credentials

Each class throws at construction time if no API key is resolved:

Error: Deepgram STT requires an apiKey. Pass { apiKey: 'dg_...' } or
set DEEPGRAM_API_KEY in the environment.

What’s Next

LLM

Configure the language model.

TTS

Configure speech synthesis.