/audio/speech

總覽

功能	支援	備註
成本追蹤	✅	可與所有支援的模型搭配使用
記錄	✅	可跨所有整合使用
端使用者追蹤	✅
備援	✅	可在支援的模型之間運作
負載平衡	✅	可在支援的模型之間運作
防護欄	✅	套用於輸入文字（僅限非串流）
支援的提供者	OpenAI, Azure OpenAI, Vertex AI, AWS Polly, ElevenLabs , MiniMax

LiteLLM Python SDK 使用方式

快速開始

from pathlib import Path
from litellm import speech
import os 

os.environ["OPENAI_API_KEY"] = "sk-.."

speech_file_path = Path(__file__).parent / "speech.mp3"
response = speech(
        model="openai/tts-1",
        voice="alloy",
        input="the quick brown fox jumped over the lazy dogs",
    )
response.stream_to_file(speech_file_path)

非同步使用方式

from litellm import aspeech
from pathlib import Path
import os, asyncio

os.environ["OPENAI_API_KEY"] = "sk-.."

async def test_async_speech(): 
    speech_file_path = Path(__file__).parent / "speech.mp3"
    response = await aspeech(
            model="openai/tts-1",
            voice="alloy",
            input="the quick brown fox jumped over the lazy dogs",
            api_base=None,
            api_key=None,
            organization=None,
            project=None,
            max_retries=1,
            timeout=600,
            client=None,
            optional_params={},
        )
    response.stream_to_file(speech_file_path)

asyncio.run(test_async_speech())

LiteLLM Proxy 使用方式

LiteLLM 提供一個與 openai 相容的 /audio/speech 端點，用於文字轉語音請求。

curl http://0.0.0.0:4000/v1/audio/speech \
  -H "Authorization: Bearer sk-1234" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tts-1",
    "input": "The quick brown fox jumped over the lazy dog.",
    "voice": "alloy"
  }' \
  --output speech.mp3

設定

- model_name: tts
  litellm_params:
    model: openai/tts-1
    api_key: os.environ/OPENAI_API_KEY

litellm --config /path/to/config.yaml

# RUNNING on http://0.0.0.0:4000

支援的提供者

提供者	使用方式連結
OpenAI	使用方式
Azure OpenAI	使用方式
Azure AI Speech Service (AVA)	使用方式
AWS Polly	使用方式
Vertex AI	使用方式
Gemini	使用方式
ElevenLabs	使用方式
MiniMax	使用方式

/audio/speech 到 /chat/completions 橋接

LiteLLM 讓您可以使用 /chat/completions 模型透過 /audio/speech 端點來產生語音。這對於像 Gemini 的已啟用 TTS 的模型特別有用，這類模型只能透過 /chat/completions 存取。

Gemini 文字轉語音

Python SDK 使用方式

Gemini Text-to-Speech SDK Usage

import litellm
import os

# Set your Gemini API key
os.environ["GEMINI_API_KEY"] = "your-gemini-api-key"

def test_audio_speech_gemini():
    result = litellm.speech(
        model="gemini/gemini-2.5-flash-preview-tts",
        input="the quick brown fox jumped over the lazy dogs",
        api_key=os.getenv("GEMINI_API_KEY"),
    )
    
    # Save to file
    from pathlib import Path
    speech_file_path = Path(__file__).parent / "gemini_speech.mp3"
    result.stream_to_file(speech_file_path)
    print(f"Audio saved to {speech_file_path}")

test_audio_speech_gemini()

非同步使用方式

Gemini Text-to-Speech Async Usage

import litellm
import asyncio
import os
from pathlib import Path

os.environ["GEMINI_API_KEY"] = "your-gemini-api-key"

async def test_async_gemini_speech():
    speech_file_path = Path(__file__).parent / "gemini_speech.mp3"
    response = await litellm.aspeech(
        model="gemini/gemini-2.5-flash-preview-tts",
        input="the quick brown fox jumped over the lazy dogs",
        api_key=os.getenv("GEMINI_API_KEY"),
    )
    response.stream_to_file(speech_file_path)
    print(f"Audio saved to {speech_file_path}")

asyncio.run(test_async_gemini_speech())

LiteLLM Proxy 使用方式

設定組態：

Gemini Proxy Configuration

model_list:
- model_name: gemini-tts
  litellm_params:
    model: gemini/gemini-2.5-flash-preview-tts
    api_key: os.environ/GEMINI_API_KEY

啟動 Proxy：

Start LiteLLM Proxy

litellm --config /path/to/config.yaml

# RUNNING on http://0.0.0.0:4000

送出請求：

Gemini TTS Request

curl http://0.0.0.0:4000/v1/audio/speech \
  -H "Authorization: Bearer sk-1234" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-tts",
    "input": "The quick brown fox jumped over the lazy dog.",
    "voice": "alloy"
  }' \
  --output gemini_speech.mp3

Vertex AI 文字轉語音

Python SDK 使用方式

Vertex AI Text-to-Speech SDK Usage

import litellm
import os
from pathlib import Path

# Set your Google credentials
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "path/to/service-account.json"

def test_audio_speech_vertex():
    result = litellm.speech(
        model="vertex_ai/gemini-2.5-flash-preview-tts",
        input="the quick brown fox jumped over the lazy dogs",
    )
    
    # Save to file
    speech_file_path = Path(__file__).parent / "vertex_speech.mp3"
    result.stream_to_file(speech_file_path)
    print(f"Audio saved to {speech_file_path}")

test_audio_speech_vertex()

LiteLLM Proxy 使用方式

設定組態：

Vertex AI Proxy Configuration

model_list:
- model_name: vertex-tts
  litellm_params:
    model: vertex_ai/gemini-2.5-flash-preview-tts
    vertex_project: your-project-id
    vertex_location: us-central1

送出請求：

Vertex AI TTS Request

curl http://0.0.0.0:4000/v1/audio/speech \
  -H "Authorization: Bearer sk-1234" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "vertex-tts",
    "input": "The quick brown fox jumped over the lazy dog.",
    "voice": "en-US-Wavenet-D"
  }' \
  --output vertex_speech.mp3

AWS Polly 文字轉語音

AWS Polly 提供神經式與標準文字轉語音引擎，支援多種語音與語言。

請參閱 AWS Polly 提供者文件 以取得詳細的使用範例。

✨ Enterprise LiteLLM Proxy - 設定最大請求檔案大小

當您想要限制送往 audio/transcriptions 的請求檔案大小時，請使用此功能

- model_name: whisper
  litellm_params:
    model: whisper-1
    api_key: sk-*******
    max_file_size_mb: 0.00001 # 👈 max file size in MB  (Set this intentionally very small for testing)
  model_info:
    mode: audio_transcription

使用有效檔案送出測試請求

curl --location 'http://localhost:4000/v1/audio/transcriptions' \
--header 'Authorization: Bearer sk-1234' \
--form 'file=@"/Users/ishaanjaffer/Github/litellm/tests/gettysburg.wav"' \
--form 'model="whisper"'

預期會看到以下回應

{"error":{"message":"File size is too large. Please check your file size. Passed file size: 0.7392807006835938 MB. Max file size: 0.0001 MB","type":"bad_request","param":"file","code":500}}%