Skip to main content
When calling POST /v1/audio/speech, send the selected voice at the request body root. Choose the model first, then select a voice supported by that model.
Voice IDs are case sensitive. Keep version suffixes and spaces exactly as shown. Voices from different models are not interchangeable; for example, longanhuan_v3.1 belongs to Qwen-Audio 3.1, not Qwen3-TTS.

Model overview

These four non-realtime models have 130 system voice entries in the source catalog. Check model availability in the Tikway model catalog. Additional Qwen-Audio 3.0 base voices are linked below.

Request examples

Use a string input and a top-level voice. You can pass a configured model ID or alias.

Qwen3-TTS

Qwen-Audio 3.1

language_type is a Qwen3-TTS extension. Put it in extra_body, or at the request body root. A root value wins if both are present. Do not send it to Qwen-Audio.

Qwen3-TTS Flash

Model: qwen3-tts-flash. Official voices and samples. General voices support Mandarin, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, and Korean. Dialect voices use the listed Chinese accent and also support the other nine languages.

General multilingual voices · 38

Dialect voices · 10

Spaces are part of the IDs Eldric Sage, Ono Anna, and Radio Gol.

Qwen-Audio 3.1 Flash

Model: qwen-audio-3.1-tts-flash. Official voices and samples. According to the provider’s regional guidance, multilingual and dialect voices and premium English voices work in Beijing and Singapore. Premium Chinese and other system voices are available in Beijing only.

Multilingual and dialect voices · 4

These four voices share support for Shanghainese, Cantonese, Northeastern, Chongqing, Shaanxi, Yunnan, Ningbo, and Gansu dialects, plus Japanese, Korean, French, German, Portuguese, Italian, Vietnamese, and Indonesian.

Premium Chinese voices · 22

For Mandarin text.

Premium English voices · 15

For English text.

Other system voices · 27

Qwen-Audio 3.0 Plus

Model: qwen-audio-3.0-tts-plus. Both system voices support Mandarin and English. Official voices and samples.

Qwen-Audio 3.0 Flash

Model: qwen-audio-3.0-tts-flash. The first nine system voices support Mandarin and English; the last three are English voices. Official voices and samples.
Version suffixes vary for 3.0 Flash: longanfengyue has no suffix, longanhuan_v3.6 must keep _v3.6, and loongjohn is all lowercase.

Base voices and samples

Qwen-Audio 3.0 Plus and Flash each offer over 500 additional base voices. These are separate from the system voice counts above. Use the full ID from the list. The qwen-audio-3.0-tts-plus-... and qwen-audio-3.0-tts-flash-... prefixes are model-specific.

Sources

Checked October 15, 2026. This page covers non-realtime HTTP speech synthesis.