text2speech
Text to Speech Conversion
Converts text into speech using various text-to-speech (TTS) engines and provides an unified interface for accessing their functionality. With this package, users can easily generate audio files of spoken words, phrases, or sentences from plain text data. The package supports multiple TTS engines, including Google's 'Cloud Text-to-Speech API', 'Amazon Polly', Microsoft's 'Cognitive Services Text to Speech REST API', the 'Speechify Text-to-Speech API', and a free TTS engine called 'Coqui TTS'.
Overview
The goal of text2speech is to harmonize various text-to-speech engines, including Amazon Polly, Coqui TTS, Google Cloud Text-to-Speech API, Microsoft Cognitive Services Text to Speech REST API, and Speechify Text-to-Speech API.
With the exception of Coqui TTS and Speechify, all these engines are accessible as R packages:
aws.pollyis a client for Amazon PollygoogleLanguageRis a client to the Google Cloud Text-to-Speech APIconradis a client to the Microsoft Cognitive Services Text to Speech REST API
You might notice Coqui TTS doesn’t have its own R package. This is because, at this time, text2speech directly incorporates the functionality of Coqui TTS.
Speechify is called directly through the Speechify Text-to-Speech
API. Create an API key at
https://platform.speechify.ai/api-keys and store it in the
SPEECHIFY_API_KEY environment variable (for example in your
.Renviron), or pass it with
tts_auth("speechify", key_or_json_file = "YOUR_KEY").
Installation
You can install this package from CRAN or the development version from GitHub with:
# Install from CRAN
install.packages("text2speech")
# or the development version from GitHub
# install.packages("devtools")
devtools::install_github("jhudsl/text2speech")
Authentication
Check for authentication. If not already authenticated, users must individually configure it for each service.
library(text2speech)
# Amazon Polly
tts_auth("amazon")
#> Warning in pollyHTTP(action = "voices", verb = "GET", query = query, ...):
#> Forbidden (HTTP 403).
#> Warning in structure(out[["Voices"]], NextToken = out[["NextToken"]]): Calling 'structure(NULL, *)' is deprecated, as NULL cannot have attributes.
#> Consider 'structure(list(), *)' instead.
#> [1] TRUE
# Coqui TTS
tts_auth("coqui")
#> [1] TRUE
# Google Cloud Text-to-Speech API
tts_auth("google")
#> [1] TRUE
# Microsoft Cognitive Services Text to Speech REST API
tts_auth("microsoft")
#> [1] TRUE
# Speechify Text-to-Speech API
tts_auth("speechify")
#> [1] TRUE
Voices
List different voice options for each service.
# Amazon Polly
voices_amazon <- tts_amazon_voices()
head(voices_amazon)
#> voice language language_code gender service
#> 1 Zeina Arabic arb Female amazon
#> 2 Zhiyu Chinese Mandarin cmn-CN Female amazon
#> 3 Naja Danish da-DK Female amazon
#> 4 Mads Danish da-DK Male amazon
#> 5 Sofie Danish da-DK Female amazon
#> 6 Ruben Dutch nl-NL Male amazon
# Coqui TTS
voices_coqui <- tts_coqui_voices()
#> Warning in system("tts --list_models", intern = TRUE): running command 'tts
#> --list_models' had status 1
#> ℹ Test out different voices on the CoquiTTS Demo (<https://huggingface.co/spaces/coqui/CoquiTTS>)
head(voices_coqui)
#> # A tibble: 0 × 5
#> # ℹ 5 variables: type <chr>, language <chr>, dataset <chr>, model_name <chr>,
#> # service <chr>
# Google Cloud Text-to-Speech API
voices_google <- tts_google_voices()
head(voices_google)
#> voice language language_code gender service
#> 1 af-ZA-Standard-A <NA> af-ZA FEMALE google
#> 2 am-ET-Standard-A <NA> am-ET FEMALE google
#> 3 am-ET-Standard-B <NA> am-ET MALE google
#> 4 am-ET-Wavenet-A <NA> am-ET FEMALE google
#> 5 am-ET-Wavenet-B <NA> am-ET MALE google
#> 6 ar-XA-Chirp3-HD-Aoede Arabic ar-XA FEMALE google
# Microsoft Cognitive Services Text to Speech REST API
voices_microsoft <- tts_microsoft_voices()
head(voices_microsoft)
#> voice
#> 1 Microsoft Server Speech Text to Speech Voice (af-ZA, AdriNeural)
#> 2 Microsoft Server Speech Text to Speech Voice (af-ZA, WillemNeural)
#> 3 Microsoft Server Speech Text to Speech Voice (am-ET, MekdesNeural)
#> 4 Microsoft Server Speech Text to Speech Voice (am-ET, AmehaNeural)
#> 5 Microsoft Server Speech Text to Speech Voice (ar-AE, FatimaNeural)
#> 6 Microsoft Server Speech Text to Speech Voice (ar-AE, HamdanNeural)
#> language language_code gender service
#> 1 Afrikaans (South Africa) af-ZA Female microsoft
#> 2 Afrikaans (South Africa) af-ZA Male microsoft
#> 3 Amharic (Ethiopia) am-ET Female microsoft
#> 4 Amharic (Ethiopia) am-ET Male microsoft
#> 5 Arabic (United Arab Emirates) ar-AE Female microsoft
#> 6 Arabic (United Arab Emirates) ar-AE Male microsoft
# Speechify Text-to-Speech API
voices_speechify <- tts_speechify_voices()
head(voices_speechify)
#> voice language language_code gender service
#> 1 aadi Hindi hi-IN male speechify
#> 2 aaliya Urdu ur-IN female speechify
#> 3 aamir Urdu ur-IN male speechify
#> 4 abhay Hindi hi-IN male speechify
#> 5 abhijit Bengali bn-IN male speechify
#> 6 abirami Tamil ta-IN female speechify
Convert text to speech
Synthesize speech with tts(text = "TEXT", service = "ENGINE")
# Amazon Polly
tts("Hello world!", service = "amazon")
# Coqui TTS
tts("Hello world!", service = "coqui")
# Google Cloud Text-to-Speech API
tts("Hello world!", service = "google")
# Microsoft Cognitive Services Text to Speech REST API
tts("Hello world!", service = "microsoft")
# Speechify Text-to-Speech API
tts("Hello world!", service = "speechify")
The resulting output will consist of a standardized tibble featuring
the following columns:
index: Sequential identifier numberoriginal_text: The text input provided by the usertext: In caseoriginal_textexceeds the character limit,textrepresents the outcome of splittingoriginal_text. Otherwise,textremains the same asoriginal_text.wav: Wave object (S4 class)file: File path to the audio fileaudio_type: The audio format, either mp3 or wavduration: The duration of the audio fileservice: The text-to-speech engine used