diff --git a/html/webpage/assets/js/script.js b/html/webpage/assets/js/script.js index 90e42a7..2d27e9d 100644 --- a/html/webpage/assets/js/script.js +++ b/html/webpage/assets/js/script.js @@ -592,6 +592,17 @@ $(document).ready(function () { } }); }) + $('#ttsgenerator').click(() => { + sidemenu.hide(); + $('#content').load('tts.html', function (response, status, xhr) { + if (status === "success") { + console.log("TTS Generator content loaded successfully"); + // pindah ke ttsgenerator.js + } else { + console.error("Error loading TTS Generator content:", xhr.status, xhr.statusText); + } + }); + }); $('#filemanagement').click(() => { sidemenu.hide(); $('#content').load('filemanagement.html', function (response, status, xhr) { diff --git a/html/webpage/assets/js/tts.js b/html/webpage/assets/js/tts.js new file mode 100644 index 0000000..f09e112 --- /dev/null +++ b/html/webpage/assets/js/tts.js @@ -0,0 +1,78 @@ +function websocket_init() { + window.addEventListener('ws_connected', () => { + console.log("tts.js ws_connected event triggered"); + }); + + window.addEventListener('ws_disconnected', () => { + console.log("tts.js ws_disconnected event triggered"); + + }); + window.addEventListener('ws_message', (event) => { + let rep = event.detail; + let cmd = rep.reply; + let data = rep.data; + if (cmd && cmd.length > 0) { + switch (cmd) { + case "start_generate_tts": + if (data && data.length>0 && "ok"===data){ + add_to_list("TTS generation started."); + $('#startstopgeneration').text("Stop Generating"); + } + break; + case "stop_generate_tts": + if (data && data.length>0 && "ok"===data){ + add_to_list("TTS generation stopped."); + $('#startstopgeneration').text("Start Generating"); + } + break; + case "tts_generate_progress": + if (data && data.length>0){ + let xx = JSON.parse(data); + if ("progress" in xx){ + if ("message" in xx){ + add_to_list(xx.message + "(" + xx.progress+" %)"); + } + } + } + } + } + }); +} + +function add_to_list(message){ + // add message to generatelogs list, with date time prefix + let li = document.createElement("li"); + let now = new Date(); + let datetime = now.toLocaleString(); + li.appendChild(document.createTextNode("[" + datetime + "] " + message)); + $('#generatelogs').append(li); +} + +$(document).ready(function () { + console.log("TTS module loaded."); + websocket_init(); + + $('#uploadjson').on('click', function () { + // TODO upload JSON file and process it + }); + $('#startstopgeneration').on('click', function () { + if ($('#startstopgeneration').text() === "Start Generating") { + // clear list generatelogs + $('#generatelogs').empty(); + let data = { + voicetype: $('#voicetype').val(), + languagetogenerate: $('#languagetogenerate').val(), + databasesource: $('#databasesource').val(), + targetas: $('#targetas').val(), + fileoperation: $('input[name="fileoperation"]:checked').val(), + autoadd: $('input[name="autoadd"]:checked').val() + } + sendCommand("start_generate_tts", JSON.stringify(data)); + } else { + sendCommand("stop_generate_tts", ""); + } + + + + }); +}); \ No newline at end of file diff --git a/html/webpage/homeadmin.html b/html/webpage/homeadmin.html index 9fdb09c..a34be78 100644 --- a/html/webpage/homeadmin.html +++ b/html/webpage/homeadmin.html @@ -88,6 +88,19 @@  Setting +
  • + + + + + + + + + + + +  Text To Speech Generator
  • diff --git a/html/webpage/tts.html b/html/webpage/tts.html new file mode 100644 index 0000000..26291dd --- /dev/null +++ b/html/webpage/tts.html @@ -0,0 +1,129 @@ + + + + + + + AAS_NewGen_28OKT25rev1 + + + + + + + + + + + + + + +
    +
    +

    Text To Speech Content Generator

    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +

    Google Parameter

    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +

    Content Generator

    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    + +
    +
    +
    +
    +
      +
      +
      +
      +
      +
      + + + + + + \ No newline at end of file diff --git a/src/codes/Somecodes.kt b/src/codes/Somecodes.kt index 0ddd63a..3964c2c 100644 --- a/src/codes/Somecodes.kt +++ b/src/codes/Somecodes.kt @@ -181,6 +181,10 @@ class Somecodes { } else cb.accept(-1,-1) } + fun GetFilename(path: String) : String { + return Path.of(path).fileName.toString() + } + fun ExtractFilesFromClassPath(resourcePath: String, outputDir: Path) { try { val resource = Somecodes::class.java.getResource(resourcePath) diff --git a/src/google/GoogleTTS.kt b/src/google/GoogleTTS.kt index 18c3b45..b822ff0 100644 --- a/src/google/GoogleTTS.kt +++ b/src/google/GoogleTTS.kt @@ -1,21 +1,126 @@ package google +import codes.Somecodes import com.google.cloud.texttospeech.v1.AudioConfig import com.google.cloud.texttospeech.v1.AudioEncoding import com.google.cloud.texttospeech.v1.SynthesisInput import com.google.cloud.texttospeech.v1.TextToSpeechClient import com.google.cloud.texttospeech.v1.VoiceSelectionParams +import content.Category +import content.Language +import content.VoiceType +import database.Soundbank import org.tinylog.Logger import java.nio.file.Files +import java.util.function.BiConsumer import kotlin.io.path.Path +import db +import java.nio.file.Paths +import kotlin.io.path.absolutePathString @Suppress("unused") class GoogleTTS(credentialJson: String = "c:/googlettsapi/gtc-cloud-aas-ae23c9552e1f.json") { + private var isGeneratingSoundbank = false init{ System.setProperty("GOOGLE_APPLICATION_CREDENTIALS", credentialJson) } - fun Speak(text: String, language : GoogleTTSLanguage, voicetype: GoogleTTSVoiceType, targetfilename: String){ + /** + * Generate soundbank + * @param voicetype Voice type, default Wavenet-A + * @param language Language of the soundbank, default INDONESIA + * @param databasesource Database source, default VOICE_1 + * @param targetas Target as, default VOICE_2 + * @param file_operation File operation, options are overwrite or skip, default overwrite + * @param auto_add Auto add, options are add or skip, default add + */ + fun GenerateSoundbank(voicetype: String = "Wavenet-A", language: Language = Language.INDONESIA, databasesource: VoiceType = VoiceType.VOICE_1, targetas: VoiceType = VoiceType.VOICE_2, file_operation: fileoperation , auto_add: autoadd, progress: BiConsumer){ + val source = db.soundDB.List + .filter { it.VoiceType==databasesource.name} + .filter { it.Language==language.name } + val target = db.soundDB.List + .filter { it.VoiceType==targetas.name} + .filter { it.Language==language.name } + if (source.isNotEmpty()){ + isGeneratingSoundbank = true + source.forEachIndexed { index, s -> + if (!isGeneratingSoundbank){ + Logger.info { "Soundbank generation stopped by user" } + progress.accept(index, "Stopped by user") + return@forEachIndexed + } + + val targetfolder = Somecodes.SoundbankDirectory(language, targetas, Category.valueOf(s.Category)) + val targetfile = Somecodes.GetFilename(s.Path) + val fullpath = targetfolder.resolve(targetfile) + + var targetrow = target.find { it.Category == s.Category && it.TAG == s.TAG } + + if (Files.exists(fullpath) && file_operation == fileoperation.SKIP){ + // file exists, skip, not creating new file + Logger.info { "Skipping existing file : ${fullpath.absolutePathString()}" } + progress.accept(index , "Skipping existing file : ${fullpath.absolutePathString()}") + if (targetrow==null && auto_add == autoadd.ADD){ + // file ada, tapi tidak ada di database target, add ke database + // add ke target database + targetrow = Soundbank( + index =0U, + Description = s.Description, + TAG = s.TAG, + Category = s.Category, + Language = s.Language, + VoiceType = targetas.name, + Path = fullpath.absolutePathString() + ) + db.soundDB.Add(targetrow) + Logger.info { "Added to target database TAG ${s.TAG} Category ${s.Category}" } + progress.accept(index , "Added to target database TAG ${s.TAG} Category ${s.Category}") + } + + } else { + // file not exists, or overwrite + if (targetrow!=null){ + // ada + Speak( + text = s.Description, + language = GoogleTTSLanguage.fromLanguage(Language.valueOf(s.Language)), + voicetype = GoogleTTSVoiceType.valueOf(voicetype), + targetfilename = fullpath.absolutePathString() + ){ + success, message -> + if (success){ + + } else { + Logger.error { "Failed to generate sound for TAG ${s.TAG} Category ${s.Category}, message : $message" } + progress.accept(index , "Failed to generate sound for TAG ${s.TAG} Category ${s.Category}, message : $message") + } + } + + } else { + // tidak ada + if (auto_add == autoadd.ADD){ + // add ke target database + } else { + // skip + Logger.info { "Skipping TAG ${s.TAG} Category ${s.Category} as it does not exist in target database" } + progress.accept(index , "Skipping TAG ${s.TAG} Category ${s.Category} as it does not exist in target database") + } + } + } + } + isGeneratingSoundbank = false + } + + } + + /** + * Stop generate soundbank + */ + fun StopGenerate(){ + isGeneratingSoundbank = false + } + + fun Speak(text: String, language : GoogleTTSLanguage, voicetype: GoogleTTSVoiceType, targetfilename: String, event: BiConsumer? = null){ TextToSpeechClient.create().use { client -> try{ val input = SynthesisInput.newBuilder().setText(text).build() @@ -32,12 +137,12 @@ class GoogleTTS(credentialJson: String = "c:/googlettsapi/gtc-cloud-aas-ae23c955 val target = Path(targetfilename) Files.write(target, bytes) Logger.info { "Speak success, file saved to : $targetfilename" } + event?.accept(true, targetfilename) } catch (e : Exception){ Logger.error { "Speak failed, message : ${e.message}" } + event?.accept(false, e.message ?: "Unknown error") } - - } } } \ No newline at end of file diff --git a/src/google/GoogleTTSLanguage.kt b/src/google/GoogleTTSLanguage.kt index c22b13a..7d6e89e 100644 --- a/src/google/GoogleTTSLanguage.kt +++ b/src/google/GoogleTTSLanguage.kt @@ -1,10 +1,27 @@ package google +import content.Language + @Suppress("unused") enum class GoogleTTSLanguage(code: String) { Indonesia("id-ID"), English("en-US"), Japanese("ja-JP"), Chinese("zh-CN"), - Arabic("ar-SA") + Arabic("ar-SA"); + + companion object { + fun fromLanguage(lang: Language) : GoogleTTSLanguage { + return when(lang) { + Language.INDONESIA -> Indonesia + Language.ENGLISH -> English + Language.JAPANESE -> Japanese + Language.CHINESE -> Chinese + Language.ARABIC -> Arabic + else -> Indonesia + } + } + } + + } \ No newline at end of file diff --git a/src/google/autoadd.kt b/src/google/autoadd.kt new file mode 100644 index 0000000..daa57ce --- /dev/null +++ b/src/google/autoadd.kt @@ -0,0 +1,7 @@ +package google + +@Suppress("unused") +enum class autoadd(name: String) { + ADD("add"), + SKIP("skip"); +} \ No newline at end of file diff --git a/src/google/fileoperation.kt b/src/google/fileoperation.kt new file mode 100644 index 0000000..94141e0 --- /dev/null +++ b/src/google/fileoperation.kt @@ -0,0 +1,7 @@ +package google + +@Suppress("unused") +enum class fileoperation(name: String) { + OVERWRITE("overwrite"), + SKIP("skip"); +} \ No newline at end of file diff --git a/src/web/WebApp.kt b/src/web/WebApp.kt index 4a15410..9761544 100644 --- a/src/web/WebApp.kt +++ b/src/web/WebApp.kt @@ -44,6 +44,7 @@ import java.time.LocalDateTime import codes.configKeys import config import database.QueueTable +import google.GoogleTTS import io.javalin.websocket.WsCloseStatus import org.tinylog.Logger import java.io.File @@ -59,6 +60,7 @@ class WebApp(val listenPort: Int, var userlist: List>, val lateinit var semiauto: Javalin val objectmapper = jacksonObjectMapper() val WsContextMap = mutableMapOf() + val ttsjob = GoogleTTS() private fun SendReply(context: WsMessageContext, command: String, value: String) { try { @@ -201,6 +203,32 @@ class WebApp(val listenPort: Int, var userlist: List>, val SendReply(wsMessageContext, cmd.command, objectmapper.writeValueAsString(reply)) } + "start_generate_tts" ->{ + val js : JsonNode = objectmapper.readTree(cmd.data) + SendReply(wsMessageContext, cmd.command, "ok") + val voicetype = js.get("voicetype")?.asText("Wavenet-A") ?: "Wavenet-A" + val languagecode = js.get("languagecode")?.asText("id-ID") ?: "id-ID" + val databasesource = js.get("databasesource")?.asText("VOICE_1") ?: "VOICE_1" + val targetas = js.get("targetas")?.asText("VOICE_2") ?: "VOICE_2" + val fileoperation = js.get("fileoperation")?.asText("overwrite") ?: "overwrite" + val autoadd = js.get("autoadd")?.asText("add") ?: "add" + Logger.info {"Starting TTS Soundbank Generation, VoiceType=$voicetype, Language"} + ttsjob.GenerateSoundbank(voicetype, languagecode, databasesource, targetas, fileoperation, autoadd){ progress, message -> + SendReply(wsMessageContext, "tts_generate_progress", objectmapper.writeValueAsString( + mapOf( + "progress" to progress, + "message" to message + ) + )) + } + } + + "stop_generate_tts" ->{ + SendReply(wsMessageContext, cmd.command, "ok") + ttsjob.StopGenerate() + + } + else -> { SendReply(wsMessageContext, cmd.command, "Unknown command") }