Alibaba Qwen-Audio 3.1 TTS Flash

Alibaba's Qwen-Audio 3.1 TTS Flash reads text aloud in 70 voices, 51 of them of one language each. About $0.002 per 1,000 characters.

Alibaba's Qwen-Audio 3.1 TTS Flash reads text aloud in 70 voices: four multilingual voices that also speak eight Chinese dialects (Shanghainese, Cantonese, Northeastern, Chongqing, Shaanxi, Yunnan, Ningbo and Gansu), 15 English voices, British and American, and 51 voices of one language each, from Bulgarian to Ukrainian. Alibaba built Flash for real-time uses such as voice assistants, conversations and customer service. Try it for a narrated video, a podcast intro, a spoken reply in an app or a lesson read in the learner's own language.

Inputs: text (required: up to 5,000 characters; tags in it such as [excited], [whispers] or [serious] set how the next sentences are read, and [laughing] or [sighing] add a sound), voice (longanhuan_v3.1 by default, one of the 4 multilingual voices; an English voice such as Emily_v3.1; or a voice of one language written like qwen-audio-3.1-tts-flash-zorza_pl for Polish), instruction (optional: how to read it, in plain words and in the language of the text, for example 'Speak slowly and calmly' or a Chinese dialect such as '请讲上海话。'), language_hints (optional: the language of the text, one code, when numbers or abbreviations come out wrong), format (mp3 by default, or wav), rate (speed, 0.5 to 2, 1 by default), pitch (0.5 to 2, 1 by default; it changes the length too, and is ignored when rate is not 1) and volume (0 to 100, 50 by default). language_hints also takes other, for the languages of the one-language voices.

It reads the text it is given in the voice you pick: a voice reads the languages of its own list, so an English voice is for English text. It does not clone or design voices, transcribe speech, translate or make music. rate and pitch change the audio after it is made and cost nothing extra; a run takes about 1.5 seconds for a sentence and about 30 seconds for 5,000 characters. Alibaba lists 139 voices: the other 69 work only in its Beijing region and are not offered here.

Price: Alibaba bills the tokens of the text and of the speech, $0.23 per 1M input tokens and $1.87 per 1M output tokens: about $0.002 for 1,000 characters of English and $0.004 for 1,000 of Chinese, an estimate from the speaking rate. 5,000 characters of English read in 6 minutes cost $0.009. A slow instruction makes longer speech and costs more. A refused or failed run costs nothing.

Price: About $0.002 per 1,000 characters of English text and $0.004 per 1,000 of Chinese, an estimate from the speaking rate: billed by the tokens of the text and the speech, $0.23 per 1M input and $1.87 per 1M output.

Run Alibaba Qwen-Audio 3.1 TTS Flash via API

curl -X POST https://api.flymy.ai/api/v1/flymyai/qwen-audio-3_1-tts-flash/predict \
  -H "X-API-KEY: $FLYMYAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text":"Welcome to FlyMy.AI. Your order has shipped and will arrive on Friday.","voice":"longanhuan_v3.1"}'

API documentation · Create a free account · Pricing

Related models