TTS API

Author: mlang

openai-hs/src/OpenAI/Client.hs

@@ -68,6 +68,11 @@ module OpenAI.Client
     AudioTranslationRequest (..),
     createTranscription,
     createAudioTranslation,
+    TextToSpeechRequest(..),
+    defaultTextToSpeechRequest,
+    TextToSpeechResponse,
+    createTextToSpeech,
+
 
     -- * Engine (deprecated)
     EngineId (..),
@@ -203,6 +208,7 @@ createAudioTranslation sc atr =
     bnd <- liftIO MP.genBoundary
     createAudioTranslationInternal sc (bnd, atr)
 
+EP1 (createTextToSpeech, TextToSpeechRequest, TextToSpeechResponse)
 EP1 (createTranscriptionInternal, (BSL.ByteString, AudioTranscriptionRequest), AudioResponseData)
 EP1 (createAudioTranslationInternal, (BSL.ByteString, AudioTranslationRequest), AudioResponseData)
 
@@ -237,7 +243,8 @@ EP2 (engineCreateEmbedding, EngineId, EngineEmbeddingCreate, (OpenAIList EngineE
              :<|> createImageVariation'
            )
     :<|> (createEmbedding')
-    :<|> ( createTranscriptionInternal'
+    :<|> (createTextToSpeech'
+             :<|> createTranscriptionInternal'
              :<|> createAudioTranslationInternal'
            )
     :<|> (createFileInternal' :<|> deleteFile')

openai-servant/openai-servant.cabal

@@ -1,6 +1,6 @@
 cabal-version: 1.12
 
--- This file has been generated from package.yaml by hpack version 0.35.2.
+-- This file has been generated from package.yaml by hpack version 0.36.0.
 --
 -- see: https://github.com/sol/hpack
 
@@ -52,6 +52,7 @@ library
     , base >=4.7 && <5
     , bytestring
     , casing
+    , http-media
     , mime-types
     , servant
     , servant-auth

openai-servant/package.yaml

@@ -27,6 +27,8 @@ dependencies:
   - time
   - vector
   - mime-types
+  - http-media
+
 
 ghc-options:
   - -Wall

openai-servant/src/OpenAI/Api.hs

@@ -1,3 +1,4 @@
+{-# LANGUAGE DataKinds #-}
 -- | The API
 module OpenAI.Api where
 
@@ -6,9 +7,24 @@ import Servant.API
 import Servant.Auth
 import Servant.Auth.Client
 import Servant.Multipart.API
+import Data.List.NonEmpty (NonEmpty((:|)))
+import Data.ByteString.Lazy (ByteString)
+import           Network.HTTP.Media ((//))
 
 type OpenAIAuth = Auth '[Bearer] ()
 
+data Audio
+
+instance Accept Audio where
+  contentTypes _ = "audio" // "mpeg" :|
+                 [ "audio" // "opus",
+                   "audio" // "aac",
+                   "audio" // "flac"
+                 ]
+
+instance MimeUnrender Audio ByteString where
+  mimeUnrender _ bs = Right bs
+
 type OpenAIApi =
   "v1" :> OpenAIApiInternal
 
@@ -46,7 +62,8 @@ type EmbeddingsApi =
   OpenAIAuth :> ReqBody '[JSON] EmbeddingCreate :> Post '[JSON] EmbeddingResponse
 
 type AudioApi =
-  OpenAIAuth :> "transcriptions" :> MultipartForm Tmp AudioTranscriptionRequest :> Post '[JSON] AudioResponseData
+  OpenAIAuth :> "speech" :> ReqBody '[JSON] TextToSpeechRequest :> Post '[Audio] TextToSpeechResponse
+    :<|> OpenAIAuth :> "transcriptions" :> MultipartForm Tmp AudioTranscriptionRequest :> Post '[JSON] AudioResponseData
     :<|> OpenAIAuth :> "translations" :> MultipartForm Tmp AudioTranslationRequest :> Post '[JSON] AudioResponseData
 
 type FilesApi =

openai-servant/src/OpenAI/Resources.hs

@@ -52,6 +52,10 @@ module OpenAI.Resources
     AudioResponseData (..),
     AudioTranscriptionRequest (..),
     AudioTranslationRequest (..),
+    TextToSpeechRequest (..),
+    defaultTextToSpeechRequest,
+    TextToSpeechResponse,
+
 
     -- * Fine tuning (out of date)
     FineTuneId (..),
@@ -507,6 +511,30 @@ data AudioResponseData = AudioResponseData
 $(deriveJSON (jsonOpts 5) ''AudioResponseData)
 
 -- | Audio create API
+
+data TextToSpeechRequest = TextToSpeechRequest
+  { ttsModel :: T.Text,
+    ttsInput :: T.Text,
+    ttsVoice :: T.Text,
+    ttsResponseFormat :: Maybe T.Text,
+    ttsSpeed :: Maybe Float
+  }
+  deriving (Show, Eq)
+
+$(deriveJSON (jsonOpts 3) ''TextToSpeechRequest)
+
+defaultTextToSpeechRequest :: T.Text -> T.Text -> T.Text -> TextToSpeechRequest
+defaultTextToSpeechRequest model voice input =
+  TextToSpeechRequest
+  { ttsModel = model,
+    ttsVoice = voice,
+    ttsInput = input,
+    ttsResponseFormat = Nothing,
+    ttsSpeed = Nothing
+  }
+
+type TextToSpeechResponse = BSL.ByteString
+
 data AudioTranscriptionRequest = AudioTranscriptionRequest
   { audtsrFile :: FilePath,
     audtsrModel :: ModelId,