feat(audio): add speech (#263)

CHANGELOG.md

@@ -5,6 +5,7 @@
 - **Chat**: add tool calls (#256)
 - **Chat**: add vision feature (#258)
 - **Config**: adding ktor engine config to support Kotlin Scripting (#261)
+- ***Audio**: add speech-to-text (#263)
 
 #### Beta
 - **Assistants**: api implementation (#259)

openai-client/src/commonMain/kotlin/com.aallam.openai.client/Audio.kt

@@ -1,9 +1,6 @@
 package com.aallam.openai.client
 
-import com.aallam.openai.api.audio.Transcription
-import com.aallam.openai.api.audio.TranscriptionRequest
-import com.aallam.openai.api.audio.Translation
-import com.aallam.openai.api.audio.TranslationRequest
+import com.aallam.openai.api.audio.*
 
 /**
  * Learn how to turn audio into text.
@@ -19,4 +16,9 @@ public interface Audio {
      * Translates audio into English.
      */
     public suspend fun translation(request: TranslationRequest): Translation
+
+    /**
+     * Generates audio from the input text.
+     */
+    public suspend fun speech(request: SpeechRequest): ByteArray
 }

openai-client/src/commonMain/kotlin/com.aallam.openai.client/internal/api/ApiPath.kt

@@ -6,6 +6,7 @@ package com.aallam.openai.client.internal.api
 internal object ApiPath {
     const val Translation = "audio/translations"
     const val Transcription = "audio/transcriptions"
+    const val Speech = "audio/speech"
     const val ChatCompletions = "chat/completions"
     const val Completions = "completions"
     const val Edits = "edits"

openai-client/src/commonMain/kotlin/com.aallam.openai.client/internal/api/AudioApi.kt

@@ -6,7 +6,9 @@ import com.aallam.openai.client.Audio
 import com.aallam.openai.client.internal.extension.appendFileSource
 import com.aallam.openai.client.internal.http.HttpRequester
 import com.aallam.openai.client.internal.http.perform
+import io.ktor.client.request.*
 import io.ktor.client.request.forms.*
+import io.ktor.http.*
 
 /**
  * Implementation of [Audio].
@@ -76,4 +78,14 @@ internal class AudioApi(val requester: HttpRequester) : Audio {
         request.responseFormat?.let { append(key = "response_format", value = it) }
         request.temperature?.let { append(key = "temperature", value = it) }
     }
+
+    override suspend fun speech(request: SpeechRequest): ByteArray {
+        return requester.perform {
+            it.post{
+                url(ApiPath.Speech)
+                setBody(request)
+                contentType(ContentType.Application.Json)
+            }
+        }
+    }
 }

openai-client/src/commonTest/kotlin/com/aallam/openai/client/TestAudio.kt

@@ -1,12 +1,12 @@
 package com.aallam.openai.client
 
-import com.aallam.openai.api.audio.AudioResponseFormat
-import com.aallam.openai.api.audio.transcriptionRequest
-import com.aallam.openai.api.audio.translationRequest
+import com.aallam.openai.api.audio.*
 import com.aallam.openai.api.file.FileSource
 import com.aallam.openai.api.model.ModelId
 import com.aallam.openai.client.internal.TestFileSystem
 import com.aallam.openai.client.internal.testFilePath
+import okio.FileSystem
+import okio.Path.Companion.toPath
 import kotlin.test.Test
 import kotlin.test.assertEquals
 import kotlin.test.assertTrue
@@ -95,4 +95,15 @@ class TestAudio : TestOpenAI() {
         assertEquals(translation.duration!!, 42.06, absoluteTolerance = 0.1)
         assertTrue { translation.segments?.isNotEmpty() ?: false }
     }
+
+    @Test
+    fun speech() = test {
+        val request = speechRequest {
+            model = ModelId("tts-1")
+            input = "The quick brown fox jumped over the lazy dog."
+            voice = Voice.Alloy
+        }
+        val audio = openAI.speech(request)
+        assertTrue { audio.isNotEmpty() }
+    }
 }

openai-client/src/jvmTest/kotlin/com/aallam/openai/client/TestChatVisionJVM.kt

@@ -40,4 +40,5 @@ class TestChatVisionJVM : TestOpenAI() {
         val content = response.choices.first().message.content.orEmpty()
         assertNotNull(content)
     }
+
 }

openai-core/src/commonMain/kotlin/com.aallam.openai.api/audio/SpeechRequest.kt

@@ -0,0 +1,87 @@
+package com.aallam.openai.api.audio
+
+import com.aallam.openai.api.OpenAIDsl
+import com.aallam.openai.api.model.ModelId
+import kotlinx.serialization.SerialName
+import kotlinx.serialization.Serializable
+
+/**
+ * Generates audio from the input text.
+ */
+@Serializable
+public data class SpeechRequest(
+
+    /**
+     * One of the available TTS models: tts-1 or tts-1-hd
+     */
+    @SerialName("model") public val model: ModelId,
+
+    /**
+     * The text to generate audio for. The maximum length is 4096 characters.
+     */
+    @SerialName("input") public val input: String,
+
+    /**
+     * The voice to use when generating the audio
+     */
+    @SerialName("voice") public val voice: Voice? = null,
+
+    /**
+     * The format to audio in.
+     */
+    @SerialName("response_format") public val responseFormat: SpeechResponseFormat? = null,
+
+    /**
+     * The speed of the generated audio. Select a value from 0.25 to 4.0. 1.0 is the default.
+     */
+    @SerialName("speed") public val speed: Double? = null,
+)
+
+/**
+ * Creates a new [SpeechRequest] instance.
+ */
+public fun speechRequest(block: SpeechRequestBuilder.() -> Unit): SpeechRequest =
+    SpeechRequestBuilder().apply(block).build()
+
+/**
+ * A speech request builder.
+ */
+@OpenAIDsl
+public class SpeechRequestBuilder {
+
+    /**
+     * One of the available TTS models: tts-1 or tts-1-hd
+     */
+    public var model: ModelId? = null
+
+    /**
+     * The text to generate audio for. The maximum length is 4096 characters.
+     */
+    public var input: String? = null
+
+    /**
+     * The voice to use when generating the audio
+     */
+    public var voice: Voice? = null
+
+    /**
+     * The format to audio in.
+     */
+    public var responseFormat: SpeechResponseFormat? = null
+
+    /**
+     * The speed of the generated audio. Select a value from 0.25 to 4.0. 1.0 is the default.
+     */
+    public var speed: Double? = null
+
+    /**
+     * Builds and returns a [SpeechRequest] instance.
+     */
+    public fun build(): SpeechRequest = SpeechRequest(
+        model = requireNotNull(model) { "model is required" },
+        input = requireNotNull(input) { "input is required" },
+        voice = voice,
+        responseFormat = responseFormat,
+        speed = speed
+    )
+}

openai-core/src/commonMain/kotlin/com.aallam.openai.api/audio/SpeechResponseFormat.kt

@@ -0,0 +1,15 @@
+package com.aallam.openai.api.audio
+
+import kotlinx.serialization.Serializable
+import kotlin.jvm.JvmInline
+
+@Serializable
+@JvmInline
+public value class SpeechResponseFormat(public val value: String) {
+    public companion object {
+        public val Mp3: SpeechResponseFormat = SpeechResponseFormat("mp3")
+        public val Opus: SpeechResponseFormat = SpeechResponseFormat("opus")
+        public val Aac: SpeechResponseFormat = SpeechResponseFormat("aac")
+        public val Flac: SpeechResponseFormat = SpeechResponseFormat("flac")
+    }
+}

openai-core/src/commonMain/kotlin/com.aallam.openai.api/audio/Voice.kt

@@ -0,0 +1,20 @@
+package com.aallam.openai.api.audio
+
+import kotlinx.serialization.Serializable
+import kotlin.jvm.JvmInline
+
+/**
+ * The voice to use when generating the audio
+ */
+@Serializable
+@JvmInline
+public value class Voice(public val value: String) {
+    public companion object {
+        public val Alloy: Voice = Voice("alloy")
+        public val Echo: Voice = Voice("echo")
+        public val Fable: Voice = Voice("fable")
+        public val Onyx: Voice = Voice("onyx")
+        public val Nova: Voice = Voice("nova")
+        public val Shimmer: Voice = Voice("shimmer")
+    }
+}