Improve API and make it returning a wave file

This commit is contained in:
2020-05-06 20:17:32 +02:00
parent c2db00d748
commit 0d75b9d9c3

View File

@@ -3,70 +3,138 @@ package io.bartek.web
import android.content.Context
import android.speech.tts.TextToSpeech
import android.speech.tts.UtteranceProgressListener
import android.util.Log
import android.widget.Toast
import fi.iki.elonen.NanoHTTPD
import fi.iki.elonen.NanoHTTPD.Response.Status.*
import io.bartek.R
import org.json.JSONObject
import java.io.BufferedInputStream
import java.io.FileInputStream
import java.io.InputStream
import java.lang.RuntimeException
import java.util.*
import kotlin.collections.HashMap
class TTSServer(port: Int, private val context: Context) : NanoHTTPD(port), TextToSpeech.OnInitListener {
data class TTSRequestData(val text: String, val language: Locale)
data class Lock(var success: Boolean = false) : Object()
data class SpeechData(val stream: InputStream, val size: Long)
class TTSProcessListener(private val uuid: String, private val lock: Lock) :
UtteranceProgressListener() {
override fun onDone(utteranceId: String?) {
if (utteranceId == uuid) {
synchronized(lock) {
lock.success = true
lock.notifyAll()
}
}
}
override fun onError(utteranceId: String?) {
if (utteranceId == uuid) {
synchronized(lock) {
lock.success = false
lock.notifyAll()
}
}
}
override fun onStart(utteranceId: String?) {}
}
class TTSServer(port: Int, private val context: Context) : NanoHTTPD(port),
TextToSpeech.OnInitListener {
private val tts = TextToSpeech(context, this)
override fun serve(session: IHTTPSession?): Response {
tts.language = Locale("pl_PL")
val uuid = UUID.randomUUID().toString()
try {
val (text, language) = getRequestData(validateRequest(session))
val (stream, size) = performTTS(text, language)
return newFixedLengthResponse(OK, "audio/x-wav", stream, size)
} catch (e: ResponseException) {
throw e
} catch (e: Exception) {
throw ResponseException(INTERNAL_ERROR, e.toString(), e)
}
}
private fun validateRequest(session: IHTTPSession?): IHTTPSession {
if (session == null) {
throw ResponseException(BAD_REQUEST, "")
}
if (session.uri != "/") {
throw ResponseException(NOT_FOUND, "")
}
if (session.method != Method.POST) {
throw ResponseException(METHOD_NOT_ALLOWED, "")
}
if (session.headers["content-type"] ?. let { it != "application/json" } != false) {
throw ResponseException(BAD_REQUEST, "")
}
return session
}
private fun performTTS(text: String, language: Locale): SpeechData {
val file = createTempFile("tmp_tts_server", ".wav")
val lock = Object()
var error = false
tts.setOnUtteranceProgressListener(object : UtteranceProgressListener() {
override fun onDone(utteranceId: String?) {
if(utteranceId == uuid) {
synchronized(lock) {
lock.notifyAll()
}
}
}
override fun onError(utteranceId: String?) {
if(utteranceId == uuid) {
error = true
synchronized(lock) {
lock.notifyAll()
}
}
}
override fun onStart(utteranceId: String?) {}
})
val uuid = UUID.randomUUID().toString()
val lock = Lock()
tts.setOnUtteranceProgressListener(TTSProcessListener(uuid, lock))
synchronized(lock) {
tts.synthesizeToFile(session?.uri, null, file, uuid)
tts.language = language
tts.synthesizeToFile(text, null, file, uuid)
lock.wait()
}
val stream = FileInputStream(file)
if (!lock.success) {
throw RuntimeException("TTS failed")
}
val stream = BufferedInputStream(FileInputStream(file))
val length = file.length()
Log.d("WAV", stream.toString())
Log.d("WAV", length.toString())
file.delete()
return when(error) {
false -> newFixedLengthResponse(Response.Status.OK, "audio/x-wav", stream, length)
else -> newFixedLengthResponse(Response.Status.INTERNAL_ERROR, "text/plain", "TTS error on $uuid")
}
return SpeechData(stream, length)
}
private fun getRequestData(session: IHTTPSession): TTSRequestData {
val map = mutableMapOf<String, String>()
session.parseBody(map)
val json = JSONObject(map["postData"] ?: "{}")
val language = Locale(json.optString("language", "en_US"))
val text = json.optString("text") ?: throw ResponseException(
BAD_REQUEST,
"The missing 'text' field is required."
)
return TTSRequestData(text, language)
}
override fun onInit(status: Int) = start()
override fun start() {
super.start()
Toast.makeText(context, context.resources.getString(R.string.server_toast_service_started), Toast.LENGTH_SHORT).show()
Toast.makeText(
context,
context.resources.getString(R.string.server_toast_service_started),
Toast.LENGTH_SHORT
).show()
}
override fun stop() {
super.stop()
Toast.makeText(context, context.resources.getString(R.string.server_toast_service_stopped), Toast.LENGTH_SHORT).show()
Toast.makeText(
context,
context.resources.getString(R.string.server_toast_service_stopped),
Toast.LENGTH_SHORT
).show()
}
}