Added phoneme padding params and pause synthesis

This commit is contained in:
Brandon Thomas committed 2016-09-02 00:20:10 -04:00
1 parent 489aff2540
commit 159eff2a1e
7 files changed
+231 -31

No files matched your search

+1
View File
@@ -1,5 +1,6 @@
// Copyright (c) 2016 Brandon Thomas <bt@brand.io, echelon@gmail.com>
pub mod pause;
pub mod speed;
pub mod volume;
+93 -11
View File
@@ -37,9 +37,12 @@ const SPEED_PARAM : &'static str = "spd";
const USE_PHONEMES_PARAM: &'static str = "up";
const USE_WORDS_PARAM: &'static str = "uw";
const VOLUME_PARAM : &'static str = "vol";
const MONOPHONE_PADDING_START_PARAM : &'static str = "mps";
const MONOPHONE_PADDING_END_PARAM : &'static str = "mpe";
const POLYPHONE_PADDING_END_PARAM : &'static str = "ppe";
/// Represents a request to this endpoint.
#[derive(Debug)]
#[derive(Clone, Debug)]
struct SpeakRequest {
/** The sentence to be spoken. */
pub sentence: String,
@@ -58,6 +61,15 @@ struct SpeakRequest {
/** Whether to use words. */
pub use_words: bool,
/** Padding before a monophone. */
pub monophone_padding_start: Option<u16>,
/** Padding after a monophone. */
pub monophone_padding_end: Option<u16>,
/** Padding after a polyphone. */
pub polyphone_padding_end: Option<u16>,
}
enum SpeakerRequestError {
@@ -82,13 +94,14 @@ pub struct AudioSynthHandler {
}
impl SpeakRequest {
pub fn parse(http_request: &mut Request) -> Result<SpeakRequest, SpeakerRequestError> {
pub fn parse(http_request: &mut Request)
-> Result<SpeakRequest, SpeakerRequestError> {
let sentence_error = Err(SpeakerRequestError::SentenceInvalid);
let speaker_error = Err(SpeakerRequestError::SpeakerInvalid);
// Get the request sentence and speaker.
// TODO: Cleanup
// TODO: OMFG WTF CLEANUP THIS GARBAGE.
match http_request.get_ref::<UrlEncodedQuery>() {
Err(_) => { return sentence_error; },
Ok(ref map) => {
@@ -179,6 +192,69 @@ impl SpeakRequest {
},
};
let mps : Option<u16> = match map.get(MONOPHONE_PADDING_START_PARAM) {
None => { None },
Some(list) => {
match list.get(0) {
None => { None },
Some(s) => {
match s.trim().parse::<u16>() {
Err(_) => None,
Ok(i) => {
if i == 0 {
None
} else {
Some(i)
}
}
}
},
}
},
};
let mpe : Option<u16> = match map.get(MONOPHONE_PADDING_END_PARAM) {
None => { None },
Some(list) => {
match list.get(0) {
None => { None },
Some(s) => {
match s.trim().parse::<u16>() {
Err(_) => None,
Ok(i) => {
if i == 0 {
None
} else {
Some(i)
}
}
}
},
}
},
};
let ppe : Option<u16> = match map.get(POLYPHONE_PADDING_END_PARAM) {
None => { None },
Some(list) => {
match list.get(0) {
None => { None },
Some(s) => {
match s.trim().parse::<u16>() {
Err(_) => None,
Ok(i) => {
if i == 0 {
None
} else {
Some(i)
}
}
}
},
}
},
};
Ok(SpeakRequest {
sentence: sen,
speaker: spk,
@@ -186,6 +262,9 @@ impl SpeakRequest {
speed: speed,
use_phonemes: use_phonemes,
use_words: use_words,
monophone_padding_start: mps,
monophone_padding_end: mpe,
polyphone_padding_end: ppe,
})
},
}
@@ -236,9 +315,7 @@ impl Handler for AudioSynthHandler {
}
}
let result = self.create_audio(&request.speaker, &request.sentence,
request.use_words, request.use_phonemes,
request.volume, request.speed);
let result = self.create_audio(request);
let mime_type = "audio/wav".parse::<Mime>().unwrap();
@@ -261,14 +338,19 @@ impl AudioSynthHandler {
// TODO: Return errors.
/// Create audio from the sentence.
fn create_audio(&self, speaker: &str, sentence: &str, use_words: bool,
use_phonemes: bool, volume: Option<f32>,
speed: Option<f32>) -> Vec<u8> {
fn create_audio(&self, request: SpeakRequest) -> Vec<u8> {
match self.synthesizer.read() {
Err(_) => Vec::new(), // TODO Actual error.
Ok(synth) => {
let generated = synth.generate(sentence, speaker, use_words,
use_phonemes, volume, speed);
let generated = synth.generate(&request.sentence,
&request.speaker,
request.use_words,
request.use_phonemes,
request.volume,
request.speed,
request.monophone_padding_start,
request.monophone_padding_end,
request.polyphone_padding_end);
match generated {
Err(e) => {
println!("Error synthesizing: {:?}", e);
+38 -11
View File
@@ -2,6 +2,7 @@
use arpabet::ArpabetDictionary;
use audiobank::Audiobank;
use effects::pause::generate_pause;
use effects::speed::change_speed;
use effects::volume::change_volume;
use error::SynthError;
@@ -41,10 +42,17 @@ impl Synthesizer {
/**
* Generate a spoken wav from text input.
*/
pub fn generate(&self, sentence: &str, speaker: &str,
use_words: bool, use_phonemes: bool,
volume: Option<f32>, speed: Option<f32>)
-> Result<WavBytes, SynthError> {
pub fn generate(&self,
sentence: &str,
speaker: &str,
use_words: bool,
use_phonemes: bool,
volume: Option<f32>,
speed: Option<f32>,
monophone_padding_start: Option<u16>,
monophone_padding_end: Option<u16>,
polyphone_padding_end: Option<u16>)
-> Result<WavBytes, SynthError> {
let mut words = split_sentence(sentence);
@@ -71,7 +79,12 @@ impl Synthesizer {
}
if !word_added && use_phonemes {
word_added = self.concatenate_polyphone(&mut concatenated_waveform, speaker, word);
word_added = self.concatenate_polyphone(&mut concatenated_waveform,
speaker,
word,
monophone_padding_start,
monophone_padding_end,
polyphone_padding_end);
}
}
@@ -115,9 +128,13 @@ impl Synthesizer {
/// Concatenate a polyphone corresponding to the word to the
/// waveform we're building. Returns whether or not the word was
/// successfully found and concatenated.
fn concatenate_polyphone(&self, concatenated_waveform: &mut Vec<i16>,
speaker: &str, word: &str) -> bool {
fn concatenate_polyphone(&self,
concatenated_waveform: &mut Vec<i16>,
speaker: &str,
word: &str,
monophone_padding_start: Option<u16>,
monophone_padding_end: Option<u16>,
polyphone_padding_end: Option<u16>) -> bool {
let polyphone = match self.arpabet_dictionary.get_polyphone(word) {
Some(p) => { p },
@@ -157,15 +174,25 @@ impl Synthesizer {
continue;
},
Some(waveform_data) => {
if monophone_padding_start.is_some() {
let pause = generate_pause(monophone_padding_start.unwrap());
concatenated_waveform.extend(pause);
}
concatenated_waveform.extend(waveform_data);
if monophone_padding_end.is_some() {
let pause = generate_pause(monophone_padding_end.unwrap());
concatenated_waveform.extend(pause);
}
},
}
}
// Insert space after polyphone.
match self.audiobank.get_misc("pause") {
None => {},
Some(pause) => { concatenated_waveform.extend(pause); },
if polyphone_padding_end.is_some() {
let pause = generate_pause(polyphone_padding_end.unwrap());
concatenated_waveform.extend(pause);
}
true
+7 -1
View File
@@ -40,17 +40,23 @@
speaker = $('select').val(),
volume = $('input#vol').val(),
speed = $('input#speed').val(),
mps = $('input#monophone_padding_start').val(),
mpe = $('input#monophone_padding_end').val(),
ppe = $('input#polyphone_padding_end').val(),
query = encodeURIComponent(sentence),
url = '/speak?v=' + speaker
+ '&s=' + query
+ '&vol=' + volume
+ '&spd=' + speed
+ '&mps=' + mps
+ '&mpe=' + mpe
+ '&ppe=' + ppe
+ '&up=' + $('#use_phonemes').prop('checked')
+ '&uw=' + $('#use_words').prop('checked');
ev.preventDefault();
Url.setState(speaker, sentence, volume, speed);
Url.setState(speaker, sentence, volume, speed, mps, mpe, ppe);
Sound.play(url);
return false;
+9 -1
View File
@@ -17,6 +17,9 @@
speaker = Url.getSpeaker() || Library.getDefaultSpeaker(),
volume = Url.getVolume() || Volume.DEFAULT,
speed = Url.getSpeed() || 1.0,
mps = Url.getMonophonePaddingStart() || 0,
mpe = Url.getMonophonePaddingEnd() || 0,
ppe = Url.getPolyphonePaddingEnd() || 600,
urlPreviousState = false;
if (speaker) {
@@ -28,12 +31,17 @@
urlPreviousState = true;
}
$('input#vol').val(volume);
speed = parseFloat(speed).toFixed(2);
$('input#speed').val(speed);
$('input#vol').val(volume);
$('#use_phonemes').prop('use_phonemes', Url.getUsePhonemes());
$('#use_words').prop('use_words', Url.getUseWords());
$('input#monophone_padding_start').val(mps);
$('input#monophone_padding_end').val(mpe);
$('input#polyphone_padding_end').val(ppe);
Library.load(function(library) { Ui.libraryLoadCallback(library); });
Sound.install();
+61 -3
View File
@@ -22,6 +22,10 @@
/** JSON key for the `use_words` param. */
_USE_WORDS_KEY: 'uw',
_MONOPHONE_PADDING_START_KEY: 'mps',
_MONOPHONE_PADDING_END_KEY: 'mpe',
_POLYPHONE_PADDING_END_KEY: 'ppe',
/** Get the sentence from the `window.location`. */
getSentence: function() {
var state = this.parseState(window.location),
@@ -84,16 +88,67 @@
}
},
/** Get the MPS value from the `window.location`. */
getMonophonePaddingStart: function() {
var state = this.parseState(window.location);
if (!state || !(this._MONOPHONE_PADDING_START_KEY in state)) {
return null;
} else {
return state[this._MONOPHONE_PADDING_START_KEY];
}
},
/** Get the MPE value from the `window.location`. */
getMonophonePaddingEnd: function() {
var state = this.parseState(window.location);
if (!state || !(this._MONOPHONE_PADDING_END_KEY in state)) {
return null;
} else {
return state[this._MONOPHONE_PADDING_END_KEY];
}
},
/** Get the PPE value from the `window.location`. */
getPolyphonePaddingEnd: function() {
var state = this.parseState(window.location);
if (!state || !(this._POLYPHONE_PADDING_END_KEY in state)) {
return null;
} else {
return state[this._POLYPHONE_PADDING_END_KEY];
}
},
// TODO: Independent functions to set both speaker and sentence separately.
/** Set the `window.history` speaker, sentence, phoneme use, and word use. */
setState: function(speaker, rawSentence, volume, speed, usePhonemes, useWords) {
var urlHash = this.fromParams(speaker, rawSentence, volume, speed, usePhonemes, useWords);
setState: function(speaker,
rawSentence,
volume,
speed,
usePhonemes,
useWords,
monophonePaddingStart,
monophonePaddingEnd,
polyphonePaddingEnd) {
var urlHash = this.fromParams(speaker, rawSentence, volume, speed,
usePhonemes, useWords, monophonePaddingStart, monophonePaddingEnd,
polyphonePaddingEnd);
console.log('set state, urlhash = ', urlHash);
window.history.replaceState(null, null, urlHash);
},
/** Encode a speaker and sentence in a URL hash. */
fromParams: function(speaker, rawSentence, volume, speed, usePhonemes, useWords) {
fromParams: function(speaker,
rawSentence,
volume,
speed,
usePhonemes,
useWords,
monophonePaddingStart,
monophonePaddingEnd,
polyphonePaddingEnd) {
var cleanedSentence = SentenceHelper.cleanSentence(rawSentence),
cleanedSpeaker = speaker, // TODO: Check against speakers.
cleanedVolume = volume, // TODO: Filter invalid values.
@@ -107,6 +162,9 @@
state[this._SPEED_KEY] = speed; // TODO: Filter invalid values.
state[this._USE_PHONEMES_KEY] = !!usePhonemes;
state[this._USE_WORDS_KEY] = !!useWords;
state[this._MONOPHONE_PADDING_START] = monophonePaddingStart;
state[this._MONOPHONE_PADDING_END] = monophonePaddingEnd;
state[this._POLYPHONE_PADDING_END] = polyphonePaddingEnd;
json = JSON.stringify(state);
+22 -4
View File
@@ -26,14 +26,32 @@
<div id="options">
<input type="checkbox" id="use_words" name="use_words"
checked="checked">
<label for="use_words">Words</label>
<label for="use_words">Use Words</label>
<input type="checkbox" id="use_phonemes" name="use_phonemes"
checked="checked">
<label for="use_phonemes">Phonemes</label>
<label for="use_phonemes">Use Phonemes</label>
</div>
<div id="more_options">
<input id="speed" value="">
<label for="speed">Speed</label>
<div>
<input id="speed" value="">
<label for="speed">Speed</label>
</div>
<div>
<input id="monophone_padding_start" value="">
<label for="monophone_padding_start">
Monophone Padding Start</label>
</div>
<div>
<input id="monophone_padding_end" value="">
<label for="monophone_padding_end">
Monophone Padding End</label>
</div>
<div>
<input id="polyphone_padding_end" value="">
<label for="polyphone_padding_end">
Polyphone Padding End</label>
</div>
</div>
</form>