WhisperCPP<T>

Models & Types · AiDotNet.SpeechRecognition.Specialized

whisper.cpp: Georgi Gerganov’s C/C++ port of Whisper, here at the base checkpoint it ships by default.

Inheritance: NeuralNetworkBase<T> → AudioNeuralNetworkBase<T> → WhisperCPP<T>

Implements: ISpeechRecognizer<T>

For Beginners

whisper.cpp listens to 16 kHz audio in 30-second windows turned into a log-mel spectrogram, encodes each window with a Transformer, and a second Transformer writes the transcript token by token while reading the encoded audio.

How It Works

whisper.cpp runs openai-whisper checkpoints converted to ggml; the network is unchanged. The defaults are the multilingual base model (Radford et al. 2022, Table 1: 6 encoder and 6 decoder layers, width 512, 8 heads, 80 mel bins). Every forward reads the 30-second window openai-whisper builds (pad_or_trim); Predict returns the decoder’s logits over the start sequence (start of transcript, language, transcribe, no timestamps), whose last row is the first token’s distribution, and Train minimizes the teacher-forced cross-entropy of a transcript given as token ids or as one distribution per token.

Example

var architecture = new NeuralNetworkArchitecture<float>(
    inputType: InputType.OneDimensional,
    taskType: NeuralNetworkTaskType.SpeechRecognition,
    inputSize: 480000, outputSize: 51865);
var model = new WhisperCPP<float>(architecture);
var text = model.Transcribe(waveform).Text;

Constructors

Constructor Summary
WhisperCPP(NeuralNetworkArchitecture<T> architecture, WhisperCPPOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
WhisperCPP(NeuralNetworkArchitecture<T> architecture, string modelPath, WhisperCPPOptions options)

Properties

Property Summary
IReadOnlyList<string> SupportedLanguages { get; }
bool SupportsStreaming { get; }
bool SupportsWordTimestamps { get; }

Methods

Method Summary
IReadOnlyDictionary<string, T> DetectLanguageProbabilities(Tensor<T> audio)
IStreamingTranscriptionSession<T> StartStreamingSession(string language)
ModelMetadata<T> GetModelMetadata()
Task<TranscriptionResult<T>> TranscribeAsync(Tensor<T> audio, string language, bool? includeTimestamps, CancellationToken cancellationToken)
Tensor<T> ForwardForTraining(Tensor<T> input)
TranscriptionResult<T> Transcribe(Tensor<T> audio, string language, bool? includeTimestamps) Transcribes 16 kHz audio with openai-whisper’s transcribe: 30-second windows, temperature fallback, and word timing when timestamps are requested.
string DetectLanguage(Tensor<T> audio)