Qwen: Qwen-Audio-3.0-TTS Plus

Qwen-Audio-3.0-TTS Plus is a text-to-speech model from Qwen, Alibaba's model family. It sits in the speech category and is described as Alibaba's higher-quality text-to-speech model, positioned above the standard tier of the same line. It generates spoken audio from text and is normally reached through the DashScope Speech Synthesizer API on Alibaba's side. It was released by the provider on 2026-07-23.

The model takes text as its only input and returns speech as its only output. There is no image, audio or video input, so it cannot clone a voice from a sample or transcribe anything; it reads written text aloud. Two voices are available, longanlingxin and longanlufeng, and you select one when you make a request. No context length is published for this model, so treat the amount of text you can send in a single call as unspecified and split long documents into smaller chunks if you need to be safe.

Because the input is text and the output is audio, the obvious uses are narration and playback. You can turn articles, documentation or knowledge base entries into listenable versions, generate voiceovers for videos and product walkthroughs, produce spoken prompts or confirmations in an application, or build accessibility features that read on-screen content to users who prefer audio. Anywhere you have text that a person would rather hear than read, this model is a candidate.

On api14 you call the model through the OpenAI-compatible API using the model id qwen/qwen-audio-3.0-tts-plus, which means existing OpenAI SDKs and clients work with a change of base URL and model string. You supply the text to be spoken and pick one of the two voices. The model exposes no tunable parameters in this listing, so there is no temperature, top_p or similar sampling control to set, and no maximum output length is published. Beyond the text and the voice choice, generation settings are fixed by the provider.

If your task is not producing speech from text, another model will fit better. Speech recognition, voice cloning from a reference clip, and audio editing are outside what this model does, since it accepts text only and emits speech only. The same applies if you need fine control over prosody through sampling parameters, because none are listed here.