ModelArchitecture type

Supported ASR model architectures for AudioTranscriber (also exported as AudioTranscriberModelArchitecture).

  • 'AUTO': Automatically detects the architecture from the model bundle (default).
  • 'OMNI': Conformer / RNN-T streaming speech recognition architecture.
  • 'TINY_GEMMA': Audio-encoder + TinyGemma decoder speech recognition architecture.

Signature:

export declare type ModelArchitecture = 'AUTO' | 'OMNI' | 'TINY_GEMMA';
export declare type AudioTranscriberModelArchitecture = ModelArchitecture;