Content validity: 2026-03 | Sources: TTS API · TTS Guide · Voice List
Text-to-speech synthesis service that produces natural, human-like voices. Supports 16+ system voices, 10 languages, streaming real-time playback, natural language instruction control for tone and emotion, and custom voices via voice cloning (from audio samples) and voice design (from text descriptions). The key differentiator is the instructions parameter, which uses natural language to precisely control speech rate, emotion, and character personality.
| Scenario | Recommended Model | Notes |
|----------|------------------|-------|
| General speech synthesis / announcements | qwen3-tts-flash | Fast, multi-language, billed per character. |
| Audiobooks / game dubbing / radio dramas | qwen3-tts-instruct-flash | Control emotion, rate, and character via instructions. |
| Brand voice customization (from text description) | qwen3-tts-vd-2026-01-26 | Design a new voice from a text description without audio samples. |
| Brand voice customization (from audio sample) | qwen3-tts-vc-2026-01-22 | Clone a voice from audio samples with high fidelity. |
| Navigation / notifications | qwen3-tts-flash | Short text, high frequency, low cost. |
import os, dashscopedashscope.base_http_api_url = 'https://dashscope.aliyuncs.com/api/v1'
resp = dashscope.MultiModalConversation.call(
model="qwen3-tts-flash",
api_key=os.getenv("DASHSCOPE_API_KEY"),
text="Today is a wonderful day to build something people love!",
voice="Cherry",
language_type="English",
)
print(resp.output.audio.url) # Audio URL, valid for 24 hours
resp = dashscope.MultiModalConversation.call(
model="qwen3-tts-flash",
api_key=os.getenv("DASHSCOPE_API_KEY"),
text="Today is a wonderful day to build something people love!",
voice="Cherry",
language_type="English",
stream=True,
)
for chunk in resp: # Each chunk contains Base64-encoded audio data
print(chunk)
Use natural language to describe the desired speech style:
resp = dashscope.MultiModalConversation.call(
model="qwen3-tts-instruct-flash",
api_key=os.getenv("DASHSCOPE_API_KEY"),
text="Welcome to our grand opening! We have amazing deals waiting for you!",
voice="Cherry",
instructions="Fast speech rate, with a clear rising intonation, suitable for introducing fashion products.",
optimize_instructions=True,
)
| voice Parameter | Gender | Description | Supported Models |
|----------------|--------|-------------|-----------------|
| Cherry | Female | Sunny, positive, friendly, and natural | Instruct + Flash + Qwen-TTS |
| Serena | Female | Gentle | Instruct + Flash + Qwen-TTS |
| Ethan | Male | Sunny, warm, energetic, and vibrant | Instruct + Flash + Qwen-TTS |
| Chelsie | Female | Two-dimensional virtual girlfriend | Instruct + Flash + Qwen-TTS |
| Momo | Female | Playful and mischievous | Instruct + Flash |
| Vivian | Female | Confident, cute, and slightly feisty | Instruct + Flash |
| Moon | Male | Bold and handsome | Instruct + Flash |
| Maia | Female | A blend of intellect and gentleness | Instruct + Flash |
| Kai | Male | A soothing audio spa for your ears | Instruct + Flash |
| Nofish | Male | A designer who cannot pronounce retroflex sounds | Instruct + Flash |
| Bella | Female | Playful little girl | Instruct + Flash |
| Jennifer | Female | Premium, cinematic-quality American English | Flash only |
| Ryan | Male | Full of rhythm, bursting with dramatic flair | Flash only |
| Katerina | Female | A mature-woman voice with rich, memorable rhythm | Flash only |
| Aiden | Male | An American English young man | Flash only |
| Eldric Sage | Male | A calm and wise elder | Flash only |
All voices support: Chinese (Mandarin), English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean.
import requestsStep 1: Upload audio sample and create a voice
url = "https://dashscope.aliyuncs.com/api/v1/services/audio/tts/customization"
resp = requests.post(url,
headers={"Authorization": f"Bearer {os.getenv('DASHSCOPE_API_KEY')}"},
files={"file": open("voice_sample.mp3", "rb")},
data={"model": "qwen-voice-enrollment", "target_model": "qwen3-tts-vc-2026-01-22",
"action": "create", "preferred_name": "my_brand_voice"})
custom_voice = resp.json()["output"]["voice"]Step 2: Use the cloned voice for synthesis
resp = dashscope.MultiModalConversation.call(
model="qwen3-tts-vc-2026-01-22", # Must match target_model
api_key=os.getenv("DASHSCOPE_API_KEY"),
text="Hello, this is a test with a cloned voice.",
voice=custom_voice,
)
| Parameter | Required | Description |
|-----------|----------|-------------|
| model | Yes | Model ID. |
| text | Yes | Text to synthesize. qwen3-tts-flash: max 600 characters. Qwen-TTS: max 512 tokens. |
| voice | Yes | System voice ID, or cloned/designed voice name. |
| language_type | No | Default: Auto. Specifying the exact language significantly improves synthesis quality over Auto. Supported: Chinese, English, Japanese, Korean, French, German, Russian, Italian, Spanish, Portuguese. |
| instructions | No | Natural language instructions for speech control. Max 1,600 tokens. Chinese and English only. Qwen3-TTS-Instruct-Flash only. |
| optimize_instructions | No | When true, the system semantically enhances instructions for better naturalness. Requires instructions to be set. |
| stream | No | false = returns audio URL. true = streams Base64-encoded audio chunks. |
1. Audio URLs are valid for only 24 hours. Download immediately after generation.
2. Specifying language_type significantly outperforms Auto. When the text is in a single language, setting the exact language improves pronunciation accuracy and naturalness.
3. instructions only works with the Instruct model. The parameter is exclusive to qwen3-tts-instruct-flash. It has no effect on other models.
4. Voice cloning target_model must match the synthesis model. Otherwise synthesis fails.
5. SpeechSynthesizer has been unified to MultiModalConversation. In the DashScope Python SDK, the old SpeechSynthesizer interface has been replaced by MultiModalConversation. Parameters are fully compatible; only the interface name needs to change.
6. Qwen-TTS (legacy) is not available in cn-beijing. Use qwen3-tts-flash or qwen3-tts-instruct-flash instead.
7. Streaming via HTTP requires the header X-DashScope-SSE: enable. The Java SDK uses the streamCall interface.
Q: How do I choose between qwen3-tts-flash and qwen3-tts-instruct-flash? A: Use flash for straightforward synthesis (announcements, navigation, reading aloud). Use instruct-flash when you need to control emotion, tone, and character expressiveness (audiobooks, dubbing). For the latest pricing comparison, see the official pricing page.
Q: How do I make synthesized speech sound more natural?
A: (1) Set language_type to match the text language. (2) Use the instruct model with instructions describing the desired style. (3) Enable optimize_instructions=True to let the system enhance the instructions.
Q: What is the difference between voice cloning and voice design? A: Cloning (VC) replicates a voice from an audio sample — suitable for reproducing an existing voice. Design (VD) creates a new voice from a text description — suitable for designing brand voices from scratch.
Q: What is the maximum text length per synthesis call? A: qwen3-tts-flash supports up to 600 characters. Qwen-TTS supports up to 512 tokens. For longer text, split into segments and concatenate the audio files.
Q: How do I achieve real-time audio playback?
A: Set stream=True. The Python SDK returns a generator; iterate over it to receive Base64-encoded audio chunks for decoding and playback.