Soniox STT

SonioxSTT streams PCM audio to the Soniox real-time transcribe WebSocket (wss://stt-rt.soniox.com/transcribe-websocket). Pure aiohttp transport, no vendor SDK required. The adapter accumulates is_final tokens into segments and flushes them when an <end> / <fin> endpoint token is received.

Install

pip install "getpatter[soniox]"

Usage

Use the namespaced import (getpatter.stt.soniox) or the flat alias (getpatter.SonioxSTT). Both auto-resolve SONIOX_API_KEY from the environment when api_key= is omitted.

# Namespaced import
from getpatter.stt import soniox

stt = soniox.STT()                                        # reads SONIOX_API_KEY
stt = soniox.STT(api_key="...", language_hints=["en", "it"])

# Flat alias (equivalent)
from getpatter import SonioxSTT

stt = SonioxSTT()

// See typescript-sdk/providers/soniox for the TS twin.

Plug it into an agent:

import asyncio
from getpatter import Patter, Twilio, SonioxSTT, ElevenLabsTTS

phone = Patter(carrier=Twilio(), phone_number="+15550001234")

agent = phone.agent(
    stt=SonioxSTT(language_hints=["en"]),                 # SONIOX_API_KEY from env
    tts=ElevenLabsTTS(voice_id="rachel"),
    system_prompt="You are a helpful assistant.",
)

asyncio.run(phone.serve(agent))

Models and rates

Soniox bills per minute of streamed audio. Default rate from getpatter.pricing:

Model	Rate / min
`stt-rt-v4` (default)	$0.002
`stt-rt-v3`	$0.002
`stt-rt-v2`	$0.002

($0.12/hr = $0.002/min. Override per-call via Patter(pricing={"soniox": {"price": ...}}).)

Languages

Soniox real-time uses language hints rather than a single language code — pass a list of BCP-47 codes that the engine considers when scoring tokens:

stt = soniox.STT(language_hints=["en", "es", "it"])
stt = soniox.STT(language_hints=["en"], language_hints_strict=True)  # reject other langs

enable_language_identification=True (default) attaches a language code to each token. Real-time STT v4 supports 60+ languages — see the Soniox language matrix.

Telephony optimization

For Twilio (mulaw 8 kHz upstream — Patter converts to PCM s16le before STT):

from getpatter.providers.soniox_stt import SonioxSTT

stt = SonioxSTT.for_twilio(api_key="...", language_hints=["en"])  # 8 kHz native

Options

Option	Default	Notes
`api_key`	`None`	Reads from `SONIOX_API_KEY` when omitted.
`model`	`"stt-rt-v4"`	`"stt-rt-v4"`, `"stt-rt-v3"`, `"stt-rt-v2"`.
`language_hints`	`None`	BCP-47 code list.
`language_hints_strict`	`False`	Restrict to the supplied hints only.
`sample_rate`	`16000`	8000, 16000, 24000 Hz.
`enable_speaker_diarization`	`False`	Server-side speaker IDs.
`enable_language_identification`	`True`	Per-token language codes.
`max_endpoint_delay_ms`	`500`	Silence (ms) before endpoint, range `[500, 3000]`.

Low-level usage

from getpatter.providers.soniox_stt import SonioxSTT

stt = SonioxSTT(api_key="...", language_hints=["en", "it"])
await stt.connect()
await stt.send_audio(pcm_chunk)                           # 16 kHz PCM s16le
async for t in stt.receive_transcripts():
    print(t.text, t.is_final, t.confidence)
await stt.close()

​Soniox STT

​Install

​Usage

​Models and rates

​Languages

​Telephony optimization

​Options

​Low-level usage

Soniox STT

Install

Usage

Models and rates

Languages

Telephony optimization

Options

Low-level usage