diff --git a/public/scripts/extensions/rvc/index.js b/public/scripts/extensions/rvc/index.js deleted file mode 100644 index b3f8bbbfe..000000000 --- a/public/scripts/extensions/rvc/index.js +++ /dev/null @@ -1,489 +0,0 @@ -/* -TODO: - - load RVC models list from extras - - Settings per characters -*/ - -import { saveSettingsDebounced } from "../../../script.js"; -import { getContext, getApiUrl, extension_settings, doExtrasFetch, ModuleWorkerWrapper, modules } from "../../extensions.js"; -export { MODULE_NAME, rvcVoiceConversion }; - -const MODULE_NAME = 'RVC'; -const DEBUG_PREFIX = " " -const UPDATE_INTERVAL = 1000 - -let charactersList = [] // Updated with module worker -let rvcModelsList = [] // Initialized only once -let rvcModelsReceived = false; - -function updateVoiceMapText() { - let voiceMapText = "" - for (let i in extension_settings.rvc.voiceMap) { - const voice_settings = extension_settings.rvc.voiceMap[i]; - voiceMapText += i + ":" - + voice_settings["modelName"] + "(" - + voice_settings["pitchExtraction"] + "," - + voice_settings["pitchOffset"] + "," - + voice_settings["indexRate"] + "," - + voice_settings["filterRadius"] + "," - + voice_settings["rmsMixRate"] + "," - + voice_settings["protect"] - + "),\n" - } - - extension_settings.rvc.voiceMapText = voiceMapText; - $('#rvc_voice_map').val(voiceMapText); - - console.debug(DEBUG_PREFIX, "Updated voice map debug text to\n", voiceMapText) -} - -//#############################// -// Extension UI and Settings // -//#############################// - -const defaultSettings = { - enabled: false, - model: "", - pitchOffset: 0, - pitchExtraction: "dio", - indexRate: 0.88, - filterRadius: 3, - rmsMixRate: 1, - protect: 0.33, - voicMapText: "", - voiceMap: {} -} - -function loadSettings() { - if (extension_settings.rvc === undefined) - extension_settings.rvc = {}; - - if (Object.keys(extension_settings.rvc).length === 0) { - Object.assign(extension_settings.rvc, defaultSettings) - } - $('#rvc_enabled').prop('checked', extension_settings.rvc.enabled); - $('#rvc_model').val(extension_settings.rvc.model); - - $('#rvc_pitch_extraction').val(extension_settings.rvc.pitchExtraction); - $('#rvc_pitch_extractiont_value').text(extension_settings.rvc.pitchExtraction); - - $('#rvc_index_rate').val(extension_settings.rvc.indexRate); - $('#rvc_index_rate_value').text(extension_settings.rvc.indexRate); - - $('#rvc_filter_radius').val(extension_settings.rvc.filterRadius); - $("#rvc_filter_radius_value").text(extension_settings.rvc.filterRadius); - - $('#rvc_pitch_offset').val(extension_settings.rvc.pitchOffset); - $('#rvc_pitch_offset_value').text(extension_settings.rvc.pitchOffset); - - $('#rvc_rms_mix_rate').val(extension_settings.rvc.rmsMixRate); - $("#rvc_rms_mix_rate_value").text(extension_settings.rvc.rmsMixRate); - - $('#rvc_protect').val(extension_settings.rvc.protect); - $("#rvc_protect_value").text(extension_settings.rvc.protect); - - $('#rvc_voice_map').val(extension_settings.rvc.voiceMapText); - -} - -async function onEnabledClick() { - extension_settings.rvc.enabled = $('#rvc_enabled').is(':checked'); - saveSettingsDebounced() -} - -async function onPitchExtractionChange() { - extension_settings.rvc.pitchExtraction = $('#rvc_pitch_extraction').val(); - saveSettingsDebounced() -} - -async function onIndexRateChange() { - extension_settings.rvc.indexRate = Number($('#rvc_index_rate').val()); - $("#rvc_index_rate_value").text(extension_settings.rvc.indexRate) - saveSettingsDebounced() -} - -async function onFilterRadiusChange() { - extension_settings.rvc.filterRadius = Number($('#rvc_filter_radius').val()); - $("#rvc_filter_radius_value").text(extension_settings.rvc.filterRadius) - saveSettingsDebounced() -} - -async function onPitchOffsetChange() { - extension_settings.rvc.pitchOffset = Number($('#rvc_pitch_offset').val()); - $("#rvc_pitch_offset_value").text(extension_settings.rvc.pitchOffset) - saveSettingsDebounced() -} - -async function onRmsMixRateChange() { - extension_settings.rvc.rmsMixRate = Number($('#rvc_rms_mix_rate').val()); - $("#rvc_rms_mix_rate_value").text(extension_settings.rvc.rmsMixRate) - saveSettingsDebounced() -} - -async function onProtectChange() { - extension_settings.rvc.protect = Number($('#rvc_protect').val()); - $("#rvc_protect_value").text(extension_settings.rvc.protect) - saveSettingsDebounced() -} - -async function onApplyClick() { - let error = false; - const character = $("#rvc_character_select").val(); - const model_name = $("#rvc_model_select").val(); - const pitchExtraction = $("#rvc_pitch_extraction").val(); - const indexRate = $("#rvc_index_rate").val(); - const filterRadius = $("#rvc_filter_radius").val(); - const pitchOffset = $("#rvc_pitch_offset").val(); - const rmsMixRate = $("#rvc_rms_mix_rate").val(); - const protect = $("#rvc_protect").val(); - - if (character === "none") { - toastr.error("Character not selected.", DEBUG_PREFIX + " voice mapping apply", { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true }); - return; - } - - if (model_name == "none") { - toastr.error("Model not selected.", DEBUG_PREFIX + " voice mapping apply", { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true }); - return; - } - - extension_settings.rvc.voiceMap[character] = { - "modelName": model_name, - "pitchExtraction": pitchExtraction, - "indexRate": indexRate, - "filterRadius": filterRadius, - "pitchOffset": pitchOffset, - "rmsMixRate": rmsMixRate, - "protect": protect - } - - updateVoiceMapText(); - - console.debug(DEBUG_PREFIX, "Updated settings of ", character, ":", extension_settings.rvc.voiceMap[character]) - saveSettingsDebounced(); -} - -async function onDeleteClick() { - const character = $("#rvc_character_select").val(); - - if (character === "none") { - toastr.error("Character not selected.", DEBUG_PREFIX + " voice mapping delete", { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true }); - return; - } - - delete extension_settings.rvc.voiceMap[character]; - console.debug(DEBUG_PREFIX, "Deleted settings of ", character); - updateVoiceMapText(); - saveSettingsDebounced(); -} - -async function onChangeUploadFiles() { - const url = new URL(getApiUrl()); - const inputFiles = $("#rvc_model_upload_files").get(0).files; - let formData = new FormData(); - - for (const file of inputFiles) - formData.append(file.name, file); - - console.debug(DEBUG_PREFIX, "Sending files:", formData); - url.pathname = '/api/voice-conversion/rvc/upload-models'; - - const apiResult = await doExtrasFetch(url, { - method: 'POST', - body: formData - }); - - if (!apiResult.ok) { - toastr.error(apiResult.statusText, DEBUG_PREFIX + ' Check extras console for errors log'); - throw new Error(`HTTP ${apiResult.status}: ${await apiResult.text()}`); - } - - alert('The files have been uploaded successfully.'); -} - -$(document).ready(function () { - function addExtensionControls() { - const settingsHtml = ` -
-
-
- RVC -
-
-
-

Characters Voice Mapping

-
- - - -
-
-
- - - -
-
- - - - - -
-
-
- - Upload one archive per model. With .pth and .index (optional) inside.
- Supported format: .zip .rar .7zip .7z - -
-
-

Model Settings

-
-
- - - - Tips: dio and pm faster, harvest slower but good.
- Torchcrepe and rmvpe are good but uses GPU. - -
-
- - - - Controls accent strength, too high may produce artifact. - -
-
- - - - Higher can reduce breathiness but may increase run time. - -
-
- - - - Recommended +12 key for male to female conversion and -12 key for female to male conversion. - -
-
- - - - Closer to 0 is closer to TTS and 1 is closer to trained voice. - Can help mask noise and sound more natural when set relatively low. - -
-
- - - - Avoid non voice sounds. Lower is more being ignored. - -
-
-
-
- -
-
-
-
- - `; - $('#extensions_settings').append(settingsHtml); - $("#rvc_enabled").on("click", onEnabledClick); - $("#rvc_voice_map").attr("disabled", "disabled");; - $('#rvc_pitch_extraction').on('change', onPitchExtractionChange); - $('#rvc_index_rate').on('input', onIndexRateChange); - $('#rvc_filter_radius').on('input', onFilterRadiusChange); - $('#rvc_pitch_offset').on('input', onPitchOffsetChange); - $('#rvc_rms_mix_rate').on('input', onRmsMixRateChange); - $('#rvc_protect').on('input', onProtectChange); - $("#rvc_apply").on("click", onApplyClick); - $("#rvc_delete").on("click", onDeleteClick); - - $("#rvc_model_upload_files").hide(); - $("#rvc_model_upload_select_button").on("click", function() {$("#rvc_model_upload_files").click()}); - - $("#rvc_model_upload_files").on("change", onChangeUploadFiles); - //$("#rvc_model_upload_button").on("click", onClickUpload); - $("#rvc_model_refresh_button").on("click", refreshVoiceList); - - } - addExtensionControls(); // No init dependencies - loadSettings(); // Depends on Extension Controls - - const wrapper = new ModuleWorkerWrapper(moduleWorker); - setInterval(wrapper.update.bind(wrapper), UPDATE_INTERVAL); - moduleWorker(); -}) - -//#############################// -// API Calls // -//#############################// - -/* - Check model installation state, return one of ["installed", "corrupted", "absent"] -*/ -async function get_models_list(model_id) { - const url = new URL(getApiUrl()); - url.pathname = '/api/voice-conversion/rvc/get-models-list'; - - const apiResult = await doExtrasFetch(url, { - method: 'POST' - }); - - if (!apiResult.ok) { - toastr.error(apiResult.statusText, DEBUG_PREFIX + ' Check model state request failed'); - throw new Error(`HTTP ${apiResult.status}: ${await apiResult.text()}`); - } - - return apiResult -} - -/* - Send an audio file to RVC to convert voice -*/ -async function rvcVoiceConversion(response, character, text) { - let apiResult - - // Check voice map - if (extension_settings.rvc.voiceMap[character] === undefined) { - //toastr.error("No model is assigned to character '"+character+"', check RVC voice map in the extension menu.", DEBUG_PREFIX+'RVC Voice map error', { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true }); - console.info(DEBUG_PREFIX, "No RVC model assign in voice map for current character " + character); - return response; - } - - const audioData = await response.blob() - if (!audioData.type in ['audio/mpeg', 'audio/wav', 'audio/x-wav', 'audio/wave', 'audio/webm']) { - throw `TTS received HTTP response with invalid data format. Expecting audio/mpeg, got ${audioData.type}` - } - console.log("Audio type received:", audioData.type) - - const voice_settings = extension_settings.rvc.voiceMap[character]; - - var requestData = new FormData(); - requestData.append('AudioFile', audioData, 'record'); - requestData.append("json", JSON.stringify({ - "modelName": voice_settings["modelName"], - "pitchExtraction": voice_settings["pitchExtraction"], - "pitchOffset": voice_settings["pitchOffset"], - "indexRate": voice_settings["indexRate"], - "filterRadius": voice_settings["filterRadius"], - "rmsMixRate": voice_settings["rmsMixRate"], - "protect": voice_settings["protect"], - "text": text - })); - - console.log("Sending tts audio data to RVC on extras server",requestData) - - const url = new URL(getApiUrl()); - url.pathname = '/api/voice-conversion/rvc/process-audio'; - - apiResult = await doExtrasFetch(url, { - method: 'POST', - body: requestData, - }); - - if (!apiResult.ok) { - toastr.error(apiResult.statusText, DEBUG_PREFIX + ' RVC Voice Conversion Failed', { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true }); - throw new Error(`HTTP ${apiResult.status}: ${await apiResult.text()}`); - } - - return apiResult; -} - -//#############################// -// Module Worker // -//#############################// - -async function refreshVoiceList() { - let result = await get_models_list(); - result = await result.json(); - rvcModelsList = result["models_list"] - - $('#rvc_model_select') - .find('option') - .remove() - .end() - .append('') - .val('none') - - for (const modelName of rvcModelsList) { - $("#rvc_model_select").append(new Option(modelName, modelName)); - } - - rvcModelsReceived = true - console.debug(DEBUG_PREFIX, "Updated model list to:", rvcModelsList); -} - -async function moduleWorker() { - updateCharactersList(); - - if (modules.includes('rvc') && !rvcModelsReceived) { - refreshVoiceList(); - } -} - -function updateCharactersList() { - let currentcharacters = new Set(); - const context = getContext(); - for (const i of context.characters) { - currentcharacters.add(i.name); - } - - currentcharacters = Array.from(currentcharacters); - currentcharacters.unshift(context.name1); - - if (JSON.stringify(charactersList) !== JSON.stringify(currentcharacters)) { - charactersList = currentcharacters - - $('#rvc_character_select') - .find('option') - .remove() - .end() - .append('') - .val('none') - - for (const charName of charactersList) { - $("#rvc_character_select").append(new Option(charName, charName)); - } - - console.debug(DEBUG_PREFIX, "Updated character list to:", charactersList); - } -} diff --git a/public/scripts/extensions/rvc/manifest.json b/public/scripts/extensions/rvc/manifest.json deleted file mode 100644 index c9b9d8b2f..000000000 --- a/public/scripts/extensions/rvc/manifest.json +++ /dev/null @@ -1,11 +0,0 @@ -{ - "display_name": "RVC", - "loading_order": 13, - "requires": ["rvc"], - "optional": [], - "js": "index.js", - "css": "style.css", - "author": "Keij#6799", - "version": "0.1.0", - "homePage": "https://github.com/SillyTavern/SillyTavern" -} diff --git a/public/scripts/extensions/rvc/style.css b/public/scripts/extensions/rvc/style.css deleted file mode 100644 index 23ef2f78d..000000000 --- a/public/scripts/extensions/rvc/style.css +++ /dev/null @@ -1,3 +0,0 @@ -.speech-toggle { - display: flex; -} diff --git a/public/scripts/extensions/tts/index.js b/public/scripts/extensions/tts/index.js index 3a3becd74..58306bcbe 100644 --- a/public/scripts/extensions/tts/index.js +++ b/public/scripts/extensions/tts/index.js @@ -8,7 +8,6 @@ import { CoquiTtsProvider } from './coqui.js' import { SystemTtsProvider } from './system.js' import { NovelTtsProvider } from './novel.js' import { power_user } from '../../power-user.js' -import { rvcVoiceConversion } from "../rvc/index.js" export { talkingAnimation }; const UPDATE_INTERVAL = 1000 @@ -415,8 +414,8 @@ async function tts(text, voiceId, char) { let response = await ttsProvider.generateTts(text, voiceId) // RVC injection - if (extension_settings.rvc.enabled) - response = await rvcVoiceConversion(response, char, text) + if (extension_settings.rvc.enabled && typeof window['rvcVoiceConversion'] === 'function') + response = await window['rvcVoiceConversion'](response, char, text) addAudioJob(response) completeTtsJob()