rixty/SillyTavern
1
0
Fork 0
mirror of https://github.com/SillyTavern/SillyTavern.git synced 2025-06-05 21:59:27 +02:00
Code Issues Packages Projects Releases Wiki Activity

IVC + support new settings and turbo model (#2067)

Browse Source
This commit is contained in:
the big boss
2024-04-11 18:29:19 +00:00
committed by GitHub
parent 334178f6f9
commit 063c81540d
1 changed files with 150 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

161
public/scripts/extensions/tts/elevenlabs.js
View File

@ -14,6 +14,8 @@ class ElevenLabsTtsProvider {
defaultSettings = { defaultSettings = {
stability: 0.75, stability: 0.75,
similarity_boost: 0.75, similarity_boost: 0.75,
style_exaggeration: 0.00,
speaker_boost: true,
apiKey: '', apiKey: '',
model: 'eleven_monolingual_v1', model: 'eleven_monolingual_v1',
voiceMap: {}, voiceMap: {},
@ -26,27 +28,57 @@ class ElevenLabsTtsProvider {
<input id="elevenlabs_tts_api_key" type="text" class="text_pole" placeholder="<API Key>"/> <input id="elevenlabs_tts_api_key" type="text" class="text_pole" placeholder="<API Key>"/>
<label for="elevenlabs_tts_model">Model</label> <label for="elevenlabs_tts_model">Model</label>
<select id="elevenlabs_tts_model" class="text_pole"> <select id="elevenlabs_tts_model" class="text_pole">
<option value="eleven_monolingual_v1">Monolingual</option> <option value="eleven_monolingual_v1">English v1</option>
<option value="eleven_multilingual_v1">Multilingual v1</option> <option value="eleven_multilingual_v1">Multilingual v1</option>
<option value="eleven_multilingual_v2">Multilingual v2</option> <option value="eleven_multilingual_v2">Multilingual v2</option>
<option value="eleven_turbo_v2">Turbo v2</option>
</select> </select>
<input id="eleven_labs_connect" class="menu_button" type="button" value="Connect" /> <input id="eleven_labs_connect" class="menu_button" type="button" value="Connect" />
<label for="elevenlabs_tts_stability">Stability: <span id="elevenlabs_tts_stability_output"></span></label> <label for="elevenlabs_tts_stability">Stability: <span id="elevenlabs_tts_stability_output"></span></label>
<input id="elevenlabs_tts_stability" type="range" value="${this.defaultSettings.stability}" min="0" max="1" step="0.05" /> <input id="elevenlabs_tts_stability" type="range" value="${this.defaultSettings.stability}" min="0" max="1" step="0.01" />
<label for="elevenlabs_tts_similarity_boost">Similarity Boost: <span id="elevenlabs_tts_similarity_boost_output"></span></label> <label for="elevenlabs_tts_similarity_boost">Similarity Boost: <span id="elevenlabs_tts_similarity_boost_output"></span></label>
<input id="elevenlabs_tts_similarity_boost" type="range" value="${this.defaultSettings.similarity_boost}" min="0" max="1" step="0.05" /> <input id="elevenlabs_tts_similarity_boost" type="range" value="${this.defaultSettings.similarity_boost}" min="0" max="1" step="0.01" />
<div id="elevenlabs_tts_v2_options" style="display: none;">
<label for="elevenlabs_tts_style_exaggeration">Style Exaggeration: <span id="elevenlabs_tts_style_exaggeration_output"></span></label>
<input id="elevenlabs_tts_style_exaggeration" type="range" value="${this.defaultSettings.style_exaggeration}" min="0" max="1" step="0.01" />
<label for="elevenlabs_tts_speaker_boost">Speaker Boost:</label>
<input id="elevenlabs_tts_speaker_boost" style="display: inline-grid" type="checkbox" />
</div>
<hr>
<div id="elevenlabs_tts_voice_cloning">
<span>Instant Voice Cloning</span><br>
<input id="elevenlabs_tts_voice_cloning_name" type="text" class="text_pole" placeholder="Voice Name"/>
<input id="elevenlabs_tts_voice_cloning_description" type="text" class="text_pole" placeholder="Voice Description"/>
<input id="elevenlabs_tts_voice_cloning_labels" type="text" class="text_pole" placeholder="Labels"/>
<div class="menu_button" id="upload_audio_file" style="width: fit-content">
<i class="fa-solid fa-file-import"></i>
<span>Upload Audio Files</span>
</div>
<input id="elevenlabs_tts_audio_files" type="file" name="audio_files" accept="audio/*" style="display: none;" multiple>
<div id="elevenlabs_tts_selected_files_list"></div>
<input id="elevenlabs_tts_clone_voice_button" class="menu_button" type="button" style="width: fit-content" value="Clone Voice">
</div>
<hr>
</div> </div>
`; `;
return html; return html;
} }
shouldInvolveExtendedSettings() {
return this.settings.model === 'eleven_multilingual_v2';
}
onSettingsChange() { onSettingsChange() {
// Update dynamically // Update dynamically
this.settings.stability = $('#elevenlabs_tts_stability').val(); this.settings.stability = $('#elevenlabs_tts_stability').val();
this.settings.similarity_boost = $('#elevenlabs_tts_similarity_boost').val(); this.settings.similarity_boost = $('#elevenlabs_tts_similarity_boost').val();
this.settings.style_exaggeration = $('#elevenlabs_tts_style_exaggeration').val();
this.settings.speaker_boost = $('#elevenlabs_tts_speaker_boost').is(':checked');
this.settings.model = $('#elevenlabs_tts_model').find(':selected').val(); this.settings.model = $('#elevenlabs_tts_model').find(':selected').val();
$('#elevenlabs_tts_stability_output').text(this.settings.stability); $('#elevenlabs_tts_stability_output').text(this.settings.stability * 100 + '%');
$('#elevenlabs_tts_similarity_boost_output').text(this.settings.similarity_boost); $('#elevenlabs_tts_similarity_boost_output').text(this.settings.similarity_boost * 100 + '%');
$('#elevenlabs_tts_style_exaggeration_output').text(this.settings.style_exaggeration * 100 + '%');
$('#elevenlabs_tts_v2_options').toggle(this.shouldInvolveExtendedSettings());
saveTtsProviderSettings(); saveTtsProviderSettings();
} }
@ -75,21 +107,28 @@ class ElevenLabsTtsProvider {
$('#elevenlabs_tts_stability').val(this.settings.stability); $('#elevenlabs_tts_stability').val(this.settings.stability);
$('#elevenlabs_tts_similarity_boost').val(this.settings.similarity_boost); $('#elevenlabs_tts_similarity_boost').val(this.settings.similarity_boost);
$('#elevenlabs_tts_style_exaggeration').val(this.settings.style_exaggeration);
$('#elevenlabs_tts_speaker_boost').prop('checked', this.settings.speaker_boost);
$('#elevenlabs_tts_api_key').val(this.settings.apiKey); $('#elevenlabs_tts_api_key').val(this.settings.apiKey);
$('#elevenlabs_tts_model').val(this.settings.model); $('#elevenlabs_tts_model').val(this.settings.model);
$('#eleven_labs_connect').on('click', () => { this.onConnectClick(); }); $('#eleven_labs_connect').on('click', () => { this.onConnectClick(); });
$('#elevenlabs_tts_similarity_boost').on('input', this.onSettingsChange.bind(this)); $('#elevenlabs_tts_similarity_boost').on('input', this.onSettingsChange.bind(this));
$('#elevenlabs_tts_stability').on('input', this.onSettingsChange.bind(this)); $('#elevenlabs_tts_stability').on('input', this.onSettingsChange.bind(this));
$('#elevenlabs_tts_style_exaggeration').on('input', this.onSettingsChange.bind(this));
$('#elevenlabs_tts_speaker_boost').on('change', this.onSettingsChange.bind(this));
$('#elevenlabs_tts_model').on('change', this.onSettingsChange.bind(this)); $('#elevenlabs_tts_model').on('change', this.onSettingsChange.bind(this));
$('#elevenlabs_tts_stability_output').text(this.settings.stability); $('#elevenlabs_tts_stability_output').text(this.settings.stability);
$('#elevenlabs_tts_similarity_boost_output').text(this.settings.similarity_boost); $('#elevenlabs_tts_similarity_boost_output').text(this.settings.similarity_boost);
$('#elevenlabs_tts_style_exaggeration_output').text(this.settings.style_exaggeration);
$('#elevenlabs_tts_v2_options').toggle(this.shouldInvolveExtendedSettings());
try { try {
await this.checkReady(); await this.checkReady();
console.debug('ElevenLabs: Settings loaded'); console.debug('ElevenLabs: Settings loaded');
} catch { } catch {
console.debug('ElevenLabs: Settings loaded, but not ready'); console.debug('ElevenLabs: Settings loaded, but not ready');
} }
this.setupVoiceCloningMenu();
} }
// Perform a simple readiness check by trying to fetch voiceIds // Perform a simple readiness check by trying to fetch voiceIds
@ -107,6 +146,63 @@ class ElevenLabsTtsProvider {
}); });
} }
setupVoiceCloningMenu() {
const audioFilesInput = document.getElementById('elevenlabs_tts_audio_files');
const selectedFilesListElement = document.getElementById('elevenlabs_tts_selected_files_list');
const cloneVoiceButton = document.getElementById('elevenlabs_tts_clone_voice_button');
const uploadAudioFileButton = document.getElementById('upload_audio_file');
const voiceCloningNameInput = document.getElementById('elevenlabs_tts_voice_cloning_name');
const voiceCloningDescriptionInput = document.getElementById('elevenlabs_tts_voice_cloning_description');
const voiceCloningLabelsInput = document.getElementById('elevenlabs_tts_voice_cloning_labels');
const updateCloneVoiceButtonVisibility = () => {
cloneVoiceButton.style.display = audioFilesInput.files.length > 0 ? 'inline-block' : 'none';
};
const clearSelectedFiles = () => {
audioFilesInput.value = '';
selectedFilesListElement.innerHTML = '';
updateCloneVoiceButtonVisibility();
};
uploadAudioFileButton.addEventListener('click', () => {
audioFilesInput.click();
});
audioFilesInput.addEventListener('change', () => {
selectedFilesListElement.innerHTML = '';
for (const file of audioFilesInput.files) {
const listItem = document.createElement('div');
listItem.textContent = file.name;
selectedFilesListElement.appendChild(listItem);
}
updateCloneVoiceButtonVisibility();
});
cloneVoiceButton.addEventListener('click', async () => {
const voiceName = voiceCloningNameInput.value.trim();
const voiceDescription = voiceCloningDescriptionInput.value.trim();
const voiceLabels = voiceCloningLabelsInput.value.trim();
if (!voiceName) {
toastr.error('Please provide a name for the cloned voice.');
return;
}
try {
await this.addVoice(voiceName, voiceDescription, voiceLabels);
toastr.success('Voice cloned successfully. Hit reload to see the new voice in the voice listing.');
clearSelectedFiles();
voiceCloningNameInput.value = '';
voiceCloningDescriptionInput.value = '';
voiceCloningLabelsInput.value = '';
} catch (error) {
toastr.error(`Failed to clone voice: ${error.message}`);
}
});
updateCloneVoiceButtonVisibility();
}
async updateApiKey() { async updateApiKey() {
// Using this call to validate API key // Using this call to validate API key
@ -206,24 +302,26 @@ class ElevenLabsTtsProvider {
async fetchTtsGeneration(text, voiceId) { async fetchTtsGeneration(text, voiceId) {
let model = this.settings.model ?? 'eleven_monolingual_v1'; let model = this.settings.model ?? 'eleven_monolingual_v1';
console.info(`Generating new TTS for voice_id ${voiceId}, model ${model}`); console.info(`Generating new TTS for voice_id ${voiceId}, model ${model}`);
const response = await fetch( const request = {
`https://api.elevenlabs.io/v1/text-to-speech/${voiceId}`,
{
method: 'POST',
headers: {
'xi-api-key': this.settings.apiKey,
'Content-Type': 'application/json',
},
body: JSON.stringify({
model_id: model, model_id: model,
text: text, text: text,
voice_settings: { voice_settings: {
stability: Number(this.settings.stability), stability: Number(this.settings.stability),
similarity_boost: Number(this.settings.similarity_boost), similarity_boost: Number(this.settings.similarity_boost),
}, },
}), };
if (this.shouldInvolveExtendedSettings()) {
request.voice_settings.style_exaggeration = Number(this.settings.style_exaggeration);
request.voice_settings.speaker_boost = Boolean(this.settings.speaker_boost);
}
const response = await fetch(`https://api.elevenlabs.io/v1/text-to-speech/${voiceId}`, {
method: 'POST',
headers: {
'xi-api-key': this.settings.apiKey,
'Content-Type': 'application/json',
}, },
); body: JSON.stringify(request),
});
if (!response.ok) { if (!response.ok) {
toastr.error(response.statusText, 'TTS Generation Failed'); toastr.error(response.statusText, 'TTS Generation Failed');
throw new Error(`HTTP ${response.status}: ${await response.text()}`); throw new Error(`HTTP ${response.status}: ${await response.text()}`);
@ -260,4 +358,33 @@ class ElevenLabsTtsProvider {
const responseJson = await response.json(); const responseJson = await response.json();
return responseJson.history; return responseJson.history;
} }
async addVoice(name, description, labels) {
const selected_files = document.querySelectorAll('input[type="file"][name="audio_files"]');
const formData = new FormData();
formData.append('name', name);
formData.append('description', description);
formData.append('labels', labels);
for (const file of selected_files) {
if (file.files.length > 0) {
formData.append('files', file.files[0]);
}
}
const response = await fetch('https://api.elevenlabs.io/v1/voices/add', {
method: 'POST',
headers: {
'xi-api-key': this.settings.apiKey,
},
body: formData,
});
if (!response.ok) {
throw new Error(`HTTP ${response.status}: ${await response.text()}`);
}
return await response.json();
}
} }