KotobaWhisperOptions
Options & Configuration · AiDotNet.SpeechRecognition.WhisperFamily
Options for Kotoba-Whisper, built on openai-whisper’s network.
Inheritance: ModelOptions → KotobaWhisperOptions
For Beginners
These options set the size of the Whisper encoder and decoder, the language and how transcription decodes.
How It Works
The defaults are Kotoba-Whisper: 32 encoder and 2 decoder layers, width 1280, 20 heads, 128 mel bins and 51866 tokens.
Constructors
| Constructor | Summary |
|---|---|
KotobaWhisperOptions() |
Initializes a new instance with default values. |
KotobaWhisperOptions(KotobaWhisperOptions other) |
Initializes a new instance by copying from another instance. |
Properties
| Property | Summary |
|---|---|
OnnxModelOptions OnnxOptions { get; set; } |
ONNX runtime options. |
bool WordTimestamps { get; set; } |
Whether transcription times every word (openai-whisper’s word_timestamps). |
double[] Temperatures { get; set; } |
Decoding temperatures tried in turn when a window’s decoding fails the compression or log-probability checks (openai-whisper’s default: 0 to 1 in steps of 0.2). |
int DecoderDim { get; set; } |
Decoder width (n_text_state): the encoder width. |
int EncoderDim { get; set; } |
Encoder width (n_audio_state). |
int MaxTextLength { get; set; } |
Decoder positions (n_text_ctx: 448). |
int NumAttentionHeads { get; set; } |
Attention heads of every encoder and decoder block (n_audio_head = n_text_head). |
int NumDecoderLayers { get; set; } |
Decoder blocks (n_text_layer). |
int NumEncoderLayers { get; set; } |
Encoder blocks (n_audio_layer). |
int NumMels { get; set; } |
Log-mel bins of the input window (80 up to large-v2, 128 from large-v3 on). |
int VocabSize { get; set; } |
Token vocabulary (n_vocab): 51,865 for the multilingual models up to large-v2, 51,866 from large-v3 on. |
string Language { get; set; } |
Language code of the audio; null detects it from the first 30 seconds. |
string ModelPath { get; set; } |
Path to an exported ONNX model, when one is used. |