Package 'carelesswhisper'

Audio sample for testing

Description

Audio sample for testing

Usage

jfk

Format

An object of class audioSample of length 176000.

---

Record audio from the default input device

Description

Record audio from the default input device

Usage

record_audio(seconds)

Arguments

seconds

recording length

Value

Numeric vector of mono sound data sampled at 16kHz

---

Perform automatic speech recognition of the given sound sample

Description

Perform automatic speech recognition of the given sound sample

Usage

whisper(ctx, snd, params = list(), verbose = FALSE, details = FALSE)

Arguments

ctx	whisper context (which you have previously created using whisper_init())
snd	Sound data. 16kHz mono audio in a numeric vector with all values in the range [-1, 1]. This package includes the function 'record_audio()' which will record audio in this format. You could also use audio::record() or any other audio package you have access to.
params	parameters for whisper. A user should usually create a default set of parameters by calling whisper_param_defaults() and then modify.
verbose	logical. be verbose? default: FALSE.
details	logical. return detailed breakdown as a data.frame? default: FALSE

Value

Character string

Examples

## Not run: 
  ctx <- whisper_init()  # Initialise the model
  snd <- record_audio(2) # record 2 seconds of audio 
  whisper(ctx, snd)      # perform speech recognition

## End(Not run)

---

Fetch a copy of the default whisper parameters

Description

n_threads

Number of threads to use when processing. Default: 4

translate

Translate from source language into english? Default: FALSE

language

language represented in audio. Use 'auto' to automatically detect language. Default: 'en'

max_len

maximum segment length in characters. Default: 0 (meaning no limit. Set to 1 to get one-word-per-segment.)

Usage

whisper_default_params()

Value

Named list of default parameters

---

Initialise whisper by loading a model

Description

Initialise whisper by loading a model

Usage

whisper_init(
  model_path = system.file("ggml-tiny.bin", package = "carelesswhisper", mustWork = TRUE),
  verbose = FALSE
)

Arguments

model_path	path to whisper.cpp model. By default this will use the "ggml-tiny.bin" file included with this package installation, which is a tiny multi-language model See README for this package, or the original whisper.cpp documentation, for how to download other models.
verbose	Be verbose about model initialisation? Logical. Default: FALSE

Value

whisper context (ctx)

---

Named list of two-letter language codes to use as language parameter

Description

Named list of two-letter language codes to use as language parameter

Usage

whisper_lang_codes

Format

An object of class list of length 99.

---