Functions — github.com/cjpais/transcribe-rs

hub / github.com/cjpais/transcribe-rs / functions

Functions374 in github.com/cjpais/transcribe-rs

⨍ Functions 374◇ Types & classes 70

↓ 62 callersMethodget

(&self, id: i64)

src/decode/tokens.rs:85

↓ 32 callersMethodfeed

( &mut self, model: &mut dyn SpeechModel, samples: &[f32], )

src/transcriber/vad_chunked.rs:240

↓ 30 callersFunctionrequire_paths

(paths: &[&Path])

tests/common/mod.rs:6

↓ 24 callersFunctionread_wav_samples

Read WAV file samples and convert them to the required format. This function reads a WAV file and converts it to the format expected by transcription

src/audio.rs:51

↓ 23 callersMethodtranscribe_file

( &self, wav_path: &std::path::Path, params: Self::RequestParams, )

src/remote/openai.rs:151

↓ 20 callersMethodas_str

(&self)

src/remote/openai.rs:135

↓ 19 callersMethodbuild

(mut self)

src/whisperfile.rs:88

↓ 19 callersFunctionmake_speech

(frame_size: usize, num_frames: usize)

src/transcriber/test_helpers.rs:83

↓ 15 callersMethodtranscribe_with

Transcribe with model-specific parameters.

src/whisperfile.rs:323

↓ 14 callersFunctionmake_silence

(frame_size: usize, num_frames: usize)

src/transcriber/test_helpers.rs:87

↓ 12 callersFunctioncreate_session

Create an ONNX session with standard settings.

src/onnx/session.rs:203

↓ 10 callersFunctionmake_smoothed

Helper: creates a SmoothedVad wrapping an EnergyVad

src/vad/mod.rs:299

↓ 10 callersMethodtranscribe

Convenience: feed all samples and finish in one call. The return value of `feed()` is intentionally discarded here — intermediate results are accumul

src/transcriber/mod.rs:145

↓ 9 callersMethodname

(self)

src/onnx/canary/mod.rs:41

↓ 8 callersMethodis_speech

(&mut self, frame: &[f32])

src/vad/mod.rs:89

↓ 8 callersFunctionread_metadata_i32

Read a custom metadata i32 value, with optional default.

src/onnx/session.rs:252

↓ 8 callersFunctionresolve_model_path

Resolve a model file path for the requested quantization level. Looks for `{name}.{suffix}.onnx` based on the quantization variant, falling back to `

src/onnx/session.rs:216

↓ 8 callersMethodtranscribe

Transcribe audio samples (16 kHz, mono, f32 in [-1, 1]). Prepends/appends silence padding based on [`TranscribeOptions`] (or engine defaults), runs [

src/lib.rs:201

↓ 7 callersMethodfeed

( &mut self, model: &mut dyn SpeechModel, samples: &[f32], )

src/transcriber/energy_adaptive_chunked.rs:177

↓ 6 callersMethoddecoder_input_name

(&self, preferred: &str, fallbacks: &[&str])

src/onnx/cohere/mod.rs:327

↓ 6 callersFunctionmerge_sequential

Merge chunk results into a single result. Text is joined with `separator` (default `" "`). Use `""` for languages that don't use space separators (Ch

src/transcriber/merge.rs:12

↓ 4 callersMethodframe_size

(&self)

src/vad/mod.rs:85

↓ 4 callersFunctionget_engine

()

tests/whisperfile.rs:44

↓ 4 callersMethodload

Load a Whisperfile engine with default server parameters.

src/whisperfile.rs:216

↓ 4 callersMethodnext_token

Given logits for the last decoder position, pick the next token. Returns `Some(token_id)` to continue decoding, or `None` to stop (EOS reached or rep

src/decode/greedy.rs:37

↓ 4 callersFunctionparse_byte_token

Parse a byte-level BPE token like `<0xE5>` into its byte value. SentencePiece tokenizers emit these for characters outside the base vocabulary (e.g.

src/decode/sentencepiece.rs:19

↓ 4 callersFunctionremove_output

( outputs: &mut ort::session::SessionOutputs, name: &str, )

src/onnx/cohere/mod.rs:402

↓ 4 callersFunctionset_ort_accelerator

Set the global ORT accelerator preference. Call once, early in the program, before any ORT models are loaded.

src/accel.rs:66

↓ 4 callersFunctionset_whisper_accelerator

Set the global whisper.cpp accelerator preference. Call once, early in the program, before any Whisper models are loaded.

src/accel.rs:191

↓ 4 callersFunctionset_whisper_gpu_device

Set the preferred GPU device index for whisper.cpp. - `-1` (default, [`GPU_DEVICE_AUTO`]): automatically select the best device (prefers dedicated GP

src/accel.rs:280

↓ 4 callersMethodtext

(mut self, name: &str, value: &str)

src/whisperfile.rs:75

↓ 3 callersMethoddrain_prefill

(&mut self)

src/vad/mod.rs:224

↓ 3 callersMethodfinish

( &mut self, model: &mut dyn SpeechModel, )

src/transcriber/vad_chunked.rs:324

↓ 3 callersMethodflush_speech_buffer

Take the entire speech buffer and transcribe it as one chunk.

src/transcriber/vad_chunked.rs:146

↓ 3 callersFunctionget_engine

()

tests/whisper.rs:34

↓ 3 callersFunctionget_ort_accelerator

Get the current ORT accelerator preference.

src/accel.rs:71

↓ 3 callersFunctionload_vocab

Load a vocabulary file where each line is `token id`. Returns a Vec indexed by token ID, and the blank token index. Replaces `▁` (U+2581) with space

src/decode/tokens.rs:9

↓ 3 callersFunctionmerge_sequential_with_separator

Merge chunk results with a custom separator between chunk texts.

src/transcriber/merge.rs:17

↓ 3 callersMethodoffset_timestamps

Shift all segment timestamps by `offset_secs`, clamping to zero. Use a negative offset to compensate for leading silence padding, or a positive offse

src/lib.rs:265

↓ 3 callersFunctionread_metadata_str

Read a custom metadata string from an ONNX session.

src/onnx/session.rs:246

↓ 3 callersMethodreset

(&mut self)

src/vad/mod.rs:101

↓ 3 callersFunctionresolve_model_file

( model_dir: &Path, candidates: &[&str], missing_name: &str, )

src/onnx/cohere/mod.rs:367

↓ 2 callersFunctionbuild_session

Internal session builder with full control over threading and EP selection.

src/onnx/session.rs:148

↓ 2 callersMethodcapabilities

(&self)

src/whisperfile.rs:487

↓ 2 callersMethodcompute_cross_kv

(&mut self, state: &mut StreamingState)

src/onnx/moonshine/streaming.rs:594

↓ 2 callersFunctioncompute_mel

Compute mel spectrogram features from audio samples. Always returns an array of shape `[num_frames, num_mels]` (time-major).

src/features/mel.rs:49

↓ 2 callersMethodcreate_state

(&self)

src/onnx/moonshine/streaming.rs:362

↓ 2 callersFunctioncreate_temp_dir

(name: &str)

src/onnx/canary/vocab.rs:139

↓ 2 callersFunctioncreate_tokens_from_timestamped_result

(timestamped_result: &TimestampedResult)

src/onnx/parakeet/mod.rs:538

↓ 2 callersFunctioncreate_word_from_tokens

(tokens: &[Token])

src/onnx/parakeet/mod.rs:595

↓ 2 callersFunctionctc_greedy_decode

CTC greedy search decoder. For each time step, selects the token with highest logit. Skips blank tokens and consecutive repeated tokens.

src/decode/ctc.rs:15

↓ 2 callersMethoddecode

(&self, tokens: &[i64])

src/onnx/moonshine/streaming.rs:246

↓ 2 callersMethoddecode

(&self, token_ids: &[i64])

src/onnx/moonshine/model.rs:404

↓ 2 callersMethoddecode_tokens

(&self, token_ids: &[i64])

src/onnx/canary/vocab.rs:116

↓ 2 callersFunctiondefault_engine

()

src/remote/openai.rs:75

↓ 2 callersMethodfile

(mut self, name: &str, filename: &str, content_type: &str, data: Vec<u8>)

src/whisperfile.rs:60

↓ 2 callersMethodfinish

( &mut self, model: &mut dyn SpeechModel, )

src/transcriber/energy_adaptive_chunked.rs:203

↓ 2 callersMethodgenerate

( &mut self, samples: &[f32], max_tokens_per_second: f32, max_tokens_override:

src/onnx/moonshine/streaming.rs:736

↓ 2 callersFunctiongroup_tokens_into_words

(tokens: &[Token])

src/onnx/parakeet/mod.rs:561

↓ 2 callersFunctionhz_to_mel

(hz: f32)

src/features/mel.rs:253

↓ 2 callersMethodinfer

( &mut self, samples: &[f32], params: &WhisperInferenceParams, )

src/whisper_cpp/mod.rs:210

↓ 2 callersMethodinfer

( &mut self, samples: &[f32], max_length: usize, )

src/onnx/moonshine/model.rs:110

↓ 2 callersMethodinfer

( &mut self, samples: &[f32], granularity: &TimestampGranularity, )

src/onnx/parakeet/mod.rs:151

↓ 2 callersMethodinfer

(&mut self, samples: &[f32])

src/onnx/gigaam/mod.rs:92

↓ 2 callersMethodinfer

( &mut self, samples: &[f32], language: &str, use_itn: bool, )

src/onnx/sense_voice/mod.rs:196

↓ 2 callersFunctionmake_window

Generate a window function of the given type and length.

src/features/mel.rs:197

↓ 2 callersFunctionmel_filterbank

Compute mel filterbank matrix of shape [num_mels, num_fft_bins].

src/features/mel.rs:209

↓ 2 callersFunctionprepend_silence

Prepend silence to audio samples. Returns a new buffer with `silence_ms` milliseconds of zeros followed by the original samples. Assumes 16 kHz sampl

src/audio.rs:104

↓ 2 callersFunctionread_metadata_float_vec

Read a comma-separated float vector from metadata.

src/onnx/session.rs:269

↓ 2 callersFunctionrequires_sequential_session

Returns true if the selected execution provider requires sequential execution and disabled memory patterns (DirectML, WebGPU).

src/onnx/session.rs:133

↓ 2 callersFunctionrms_energy

Compute RMS energy of an audio frame.

src/transcriber/mod.rs:51

↓ 2 callersMethodtoken_rate

(&self)

src/onnx/moonshine/mod.rs:54

↓ 2 callersMethodtranscribe_chunk

Transcribe a chunk of audio samples at a given position. Does not touch `self.speech_buffer` or `self.speech_start_sample` — callers manage buffer sta

src/transcriber/vad_chunked.rs:163

↓ 2 callersMethodtranscribe_chunk

( &mut self, model: &mut dyn SpeechModel, chunk: &[f32], chunk_start_samples:

src/transcriber/energy_adaptive_chunked.rs:110

↓ 2 callersFunctiontranscribe_padded

Transcribe a chunk with optional silence padding and timestamp adjustment. Sets leading/trailing silence on [`TranscribeOptions`] and delegates paddi

src/transcriber/mod.rs:68

↓ 2 callersMethodtranscribe_raw

( &mut self, samples: &[f32], options: &TranscribeOptions, )

src/whisperfile.rs:491

↓ 2 callersMethodtranscribe_samples_inner

( &self, samples: &[f32], params: Option<WhisperfileInferenceParams>, )

src/whisperfile.rs:377

↓ 2 callersMethodtranscribe_wav_bytes

( &self, wav_data: Vec<u8>, params: Option<WhisperfileInferenceParams>, )

src/whisperfile.rs:403

↓ 2 callersMethodwith_max_repeats

(mut self, n: usize)

src/decode/greedy.rs:28

↓ 1 callersMethodapply_base64_decode

(&mut self)

src/decode/tokens.rs:74

↓ 1 callersFunctionapply_cmvn

Apply Cepstral Mean-Variance Normalization (CMVN). Formula: `x[i] = (x[i] + neg_mean[i]) * inv_stddev[i]` Modifies features in-place.

src/features/cmvn.rs:8

↓ 1 callersFunctionapply_lfr

Apply Lower Frame Rate (LFR) stacking. Concatenates `window_size` consecutive frames with a stride of `window_shift`, reducing temporal resolution wh

src/features/lfr.rs:10

↓ 1 callersFunctionargmax

(logits: &[f32])

src/decode/greedy.rs:63

↓ 1 callersMethodas_arg

(&self)

src/whisperfile.rs:146

↓ 1 callersFunctionauto_select_gpu_device

Auto-select the best GPU device for whisper inference. Prefers dedicated GPUs over integrated, then picks the device with the most total VRAM as a ti

src/whisper_cpp/gpu.rs:111

↓ 1 callersFunctionbinary_path

()

tests/whisperfile.rs:9

↓ 1 callersMethodbuild_prompt

( &self, src_lang: &str, tgt_lang: &str, use_pnc: bool, use_itn: bool,

src/onnx/canary/vocab.rs:82

↓ 1 callersMethodbuild_prompt_ids

(&self, language: Option<&str>)

src/onnx/cohere/mod.rs:260

↓ 1 callersFunctioncompute_fbank

FBANK-style feature extraction (SenseVoice): pre-emphasis + windowed frames + mel filterbank. Returns [num_frames, num_mels].

src/features/mel.rs:62

↓ 1 callersFunctioncompute_mel_spectrogram

Standard mel spectrogram (GigaAM-style): windowed STFT + mel filterbank + log. Returns [num_frames, num_mels].

src/features/mel.rs:151

↓ 1 callersMethodconvert_result

(&self, decoder_result: &CtcDecoderResult)

src/onnx/sense_voice/mod.rs:357

↓ 1 callersFunctionconvert_timestamps

( timestamped_result: &TimestampedResult, granularity: &TimestampGranularity, )

src/onnx/parakeet/mod.rs:455

↓ 1 callersFunctionconvert_to_hierarchical_segment_segments

( timestamped_result: &TimestampedResult, )

src/onnx/parakeet/mod.rs:516

↓ 1 callersFunctionconvert_to_hierarchical_word_segments

( timestamped_result: &TimestampedResult, )

src/onnx/parakeet/mod.rs:495

↓ 1 callersFunctionconvert_to_raw_token_segments

( timestamped_result: &TimestampedResult, )

src/onnx/parakeet/mod.rs:468

↓ 1 callersMethodcreate_decoder_state

(&self)

src/onnx/parakeet/mod.rs:219

↓ 1 callersFunctioncreate_session_with_threads

Create an ONNX session with configurable thread count.

src/onnx/session.rs:208

↓ 1 callersFunctiondecode_autoregressive

( decoder: &mut Session, encoder_embeddings: &DynValue, encoder_mask: &DynValue, prompt_tokens

src/onnx/canary/decoder.rs:10

↓ 1 callersMethoddecode_ids

(&self, token_ids: &[i64])

src/onnx/cohere/mod.rs:296

↓ 1 callersMethoddecode_sequence

( &mut self, encodings: &ArrayViewD<f32>, encodings_len: usize, )

src/onnx/parakeet/mod.rs:328

next →1–100 of 374, ranked by callers