feat(openai): add new TTS model and voices, include optional instruct… (#351)

Co-authored-by: David Zhao <dz@livekit.io>

Author: tomc98

.changeset/selfish-ducks-compare.md

@@ -0,0 +1,5 @@
+---
+"@livekit/agents-plugin-openai": patch
+---
+
+feat(openai): add new TTS model and voices, include optional instruct…

plugins/openai/package.json

@@ -40,7 +40,7 @@
     "typescript": "^5.0.0"
   },
   "dependencies": {
-    "openai": "^4.70.2",
+    "openai": "^4.91.1",
     "sharp": "^0.33.5",
     "ws": "^8.16.0"
   },

plugins/openai/src/models.ts

@@ -29,9 +29,20 @@ export type ChatModels =
 
 export type WhisperModels = 'whisper-1';
 
-export type TTSModels = 'tts-1' | 'tts-1-hd';
+export type TTSModels = 'tts-1' | 'tts-1-hd' | 'gpt-4o-mini-tts';
 
-export type TTSVoices = 'alloy' | 'echo' | 'fable' | 'onyx' | 'nova' | 'shimmer';
+export type TTSVoices =
+  | 'alloy'
+  | 'ash'
+  | 'ballad'
+  | 'coral'
+  | 'echo'
+  | 'fable'
+  | 'nova'
+  | 'onyx'
+  | 'sage'
+  | 'shimmer'
+  | 'verse';
 
 // adapters for OpenAI-compatible LLMs, TTSs, STTs
 

plugins/openai/src/tts.ts

@@ -14,6 +14,7 @@ export interface TTSOptions {
   model: TTSModels | string;
   voice: TTSVoices;
   speed: number;
+  instructions?: string;
   baseURL?: string;
   client?: OpenAI;
   apiKey?: string;
@@ -66,6 +67,7 @@ export class TTS extends tts.TTS {
         input: text,
         model: this.#opts.model,
         voice: this.#opts.voice,
+        instructions: this.#opts.instructions,
         response_format: 'pcm',
         speed: this.#opts.speed,
       }),