STT

Entry point group: opm.stt Config entry point group: opm.stt.config Template: ovos_plugin_manager.templates.stt Factory: ovos_plugin_manager.stt.OVOSSTTFactory

---

Module-level helpers (ovos_plugin_manager.stt)

find_stt_plugins() -> dict

Return a dict of {entry_point_name: class} for all installed STT plugins.

load_stt_plugin(module_name: str) -> Type[STT]

Return the uninstantiated class for the named plugin.

get_stt_configs() -> dict

Return {plugin_name: {lang: [config_dicts]}} for all installed STT plugins.

get_stt_module_configs(module_name: str) -> dict

Return {lang: [config_dicts]} for a single plugin, sorted by priority.

get_stt_lang_configs(lang: str, include_dialects: bool = False) -> dict

Return {plugin_name: [sorted_configs]} for all plugins that support lang.

get_stt_supported_langs() -> dict

Return {lang: [plugin_names]} for all installed STT plugins.

get_stt_config(config: dict = None, module: str = None) -> dict

Resolve merged STT config for the given module from the global configuration. Equivalent to get_plugin_config(config, "stt", module). Raises AssertionError if lang is not present in the resolved config.

---

OVOSSTTFactory

from ovos_plugin_manager.stt import OVOSSTTFactory

OVOSSTTFactory.get_class(config=None) -> Type[STT]

Return the STT class specified by config['module'].

OVOSSTTFactory.create(config=None) -> STT

Instantiate and return an STT engine based on config (or the global config). Raises on failure with a descriptive log.

---

STT base class

from ovos_plugin_manager.templates.stt import STT

Constructor

STT(config: Optional[dict] = None)

Attribute	Type	Description
config	dict	Plugin-specific configuration dict.
can_stream	bool	True if this plugin supports streaming (set by StreamingSTT).
lang	str	Active language (BCP-47). Reads from config, then from the current Session.

Abstract methods (must implement)

execute(audio: AudioData, language: Optional[str] = None) -> str

Transcribe audio and return the best single transcription string.

available_languages: Set[str] (classproperty)

Return the set of BCP-47 language codes supported by this plugin.

Provided methods (may override)

transcribe(audio: AudioData, lang: Optional[str] = None) -> List[Tuple[str, float]]

Return a list of (transcription, confidence) pairs. The default implementation wraps execute and returns [(text, 1.0)].

If lang="auto", language detection is attempted via the bound AudioLanguageDetector before calling execute.

detect_language(audio, valid_langs=None) -> Tuple[str, float]

Run language detection using the bound AudioLanguageDetector. Requires bind() to have been called first. Raises NotImplementedError if no detector is bound.

bind(detector: AudioLanguageDetector)

Attach an AudioLanguageDetector for use in detect_language / auto-lang transcribe.

runtime_requirements (classproperty) -> RuntimeRequirements

Override to declare connectivity needs. Defaults to RuntimeRequirements() (internet not required).

---

StreamingSTT base class

Extends STT. Use when audio is fed incrementally as it arrives.

from ovos_plugin_manager.templates.stt import StreamingSTT, StreamThread

Constructor

StreamingSTT(config: Optional[dict] = None)

Sets can_stream = True and creates transcript_ready: Event.

Methods

stream_start(language: Optional[str] = None)

Start a new streaming session. Stops any existing stream, creates a new Queue and StreamThread, and starts the thread.

stream_data(data: bytes)

Feed a raw audio chunk to the active stream thread's queue.

stream_stop() -> Optional[str]

Signal end-of-stream, wait for the thread to finish, and return the final transcription.

create_streaming_thread() -> StreamThread (abstract)

Return a new StreamThread instance. Implementations must create a thread that reads from self.queue and stores its result in self.stream.text.

---

StreamThread base class

from ovos_plugin_manager.templates.stt import StreamThread

ABC for the worker thread used with StreamingSTT. Subclass and implement handle_audio_stream.

handle_audio_stream(audio: Generator, language: str) (abstract)

Process the audio stream and store the final transcription in self.text.

---

Configuration

STT config is read from ovos.conf under the stt key:

{
  "stt": {
    "module": "ovos-stt-plugin-whisper",
    "lang": "en-US",
    "ovos-stt-plugin-whisper": {
      "model": "base"
    }
  }
}