Google Cloud STT

Google Cloud Speech-to-Text is a speech-to-text plugin. It transcribes the caller’s audio into text for the LLM using Google Cloud’s Speech-to-Text V2 streaming API.

Setup

Google Cloud STT authenticates with a Google Cloud service account rather than a simple API key. Point Application Default Credentials at your service-account JSON. Create the service account and key in the Google Cloud console:

export GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json

In the Python SDK you can also pass the service-account JSON path (or raw JSON string) as the api_key argument. The value is used directly as the service-account credentials; it is not written to the GOOGLE_APPLICATION_CREDENTIALS environment variable.

Usage

Import the plugin and pass it to the pipeline’s stt slot.

from zrt.plugins import GoogleSTT

stt = GoogleSTT(
    languages="en-US",
    model="latest_long",
)

# Pipeline(stt=stt, ...)

Parameters

Constructor parameters for the Python SDK (GoogleSTT). The JavaScript and Go SDKs expose equivalent options where supported, in their idiomatic form (camelCase / STTOptions struct).

Parameter	Type	Default	Description
`credentials_json`	`str`	`None`	Service-account credentials as a raw JSON string or a path to a JSON key file. Takes highest precedence.
`service_account_path`	`str`	`None`	Path to a service-account JSON key file. Used when `credentials_json` is not set.
`api_key`	`str`	`None`	Fallback credential source; the value is treated as a JSON file path (or raw JSON) first. Used only when the above are absent. Service-account auth is recommended for STT.
`project_id`	`str`	`None`	Google Cloud project ID. Extracted from the service-account JSON when not set.
`languages`	`str \| list[str]`	`"en-US"`	Language code(s) to recognize. The first entry is the primary language.
`model`	`str`	`"latest_long"`	Google Cloud Speech recognition model.
`sample_rate`	`int`	`48000`	Target sample rate (Hz). Matches the native input rate so no resampling occurs; set a lower value (e.g. `16000`) to downsample.
`interim_results`	`bool`	`True`	Emit partial transcripts while the caller is still speaking.
`punctuate`	`bool`	`True`	Add automatic punctuation to transcripts.
`min_confidence_threshold`	`float`	`0.0`	Drop results below this confidence. `0.0` disables filtering.
`location`	`str`	`"us"`	Cloud Speech-to-Text regional endpoint: `"us"` or `"eu"`.
`profanity_filter`	`bool`	`False`	Mask profane words in transcripts.
`enable_voice_activity_events`	`bool`	`False`	Emit server-side speech-start/-end events independently of transcripts.
`speech_start_timeout`	`float`	`None`	Seconds to wait for speech to begin before a voice-activity timeout fires. `None` uses the API default.
`speech_end_timeout`	`float`	`None`	Seconds of silence after speech before an end-of-speech event fires. `None` uses the API default.
`audio_channel_count`	`int`	`1`	Number of audio channels in the input (mono by default).
`min_speaker_count`	`int`	`None`	Minimum expected speakers for diarization. `None` disables diarization.
`max_speaker_count`	`int`	`None`	Maximum expected speakers for diarization. `None` disables diarization.

Import paths

SDK	Import	Constructor
Python	`from zrt.plugins import GoogleSTT`	`GoogleSTT(...)`
JavaScript	`from '@zrt/js-sdk/plugins/google'`	`GoogleSTT({ ... })`
Go	`.../zrt-golang-sdk/plugins/google`	`google.NewSTT(google.STTOptions{...})`

Transcribed text passes to the LLM once turn detection decides the caller has finished speaking.

Plugins

Inference Gateway

Speech-to-text

LLM

Text-to-speech

Realtime models

Avatars

Turn detection

Voice activity detection

Noise cancellation

Setup

Usage

Parameters

Import paths

​Setup

​Usage

​Parameters

​Import paths

Setup

Usage

Parameters

Import paths