Crate vosk

Safe FFI bindings around the Vosk API Speech Recognition Toolkit.

Basic usage:

• Create a Model
• Create a Recognizer with that model
• Feel audio to the recognizer with Recognizer::accept_waveform
• Get the processed result with Recognizer::result, Recognizer::partial_result or Recognizer::final_result

Structs

Alternative

An alternative transcript in a CompleteResultMultiple.

BatchModelbatch

The same as Model, but uses a CUDA enabled Nvidia GPU and dynamic batching to enable higher throughput.

BatchRecognizerbatch

The main object which processes data using GPU inferencing. Takes audio as input and returns decoded information as words, confidences, times, and other metadata.

CompleteResultMultiple

Recognition result if Recognizer::set_max_alternatives is passed a non-zero value.

CompleteResultSingle

Recognition result if Recognizer::set_max_alternatives is passed a zero (default).

Model

Model that stores all the data required for recognition.

PartialResult

Result returned by Recognizer::partial_result. The result may change after processing more data as decoding is not yet complete.

Recognizer

The main object which processes data. Takes audio as input and returns decoded information as words, confidences, times, and other metadata.

SpeakerInfo

Data useful for speaker identification.

SpeakerModel

The same as Model but contains the data for speaker identification.

Word

A single word in a CompleteResultSingle and metadata about it.

WordInAlternative

A single word in an Alternative and metadata about it.

Enums

AcceptWaveformError

Possible errors that accept_waveform methods might return.

CompleteResult

Different results that can be returned from Recognizer::result and Recognizer::final_result.

DecodingState

State of the decodification after processing a chunk of data.

LogLevel

Log level for Kaldi messages.

Functions

gpu_initbatch

Init, automatically select a CUDA device and allow multithreading. Must be called once from the main thread.

gpu_thread_initbatch

Init CUDA device in a multi-threaded environment. Must be called for each thread.

set_log_level

Set log level for Kaldi messages.