Qwen-Audio-3.0-TTS Flash Multilingual Speech Synthesis Example

This piece uses the qwen-audio-3.0-tts-flash author route to genuinely generate a narration that mixes Chinese and English, demonstrating the actual input parameters and output of this model in a hosted TTS scenario. The voice uses the Flash tier's built-in timbre longanhuan_v3.6, and language_hints takes only the first item, so this example is mainly Chinese with embedded English terms, precisely covering the boundary behavior of language hints. In base, rate, pitch, volume, sample_rate, format, and seed are all explicitly written out, making it convenient for developers to check the meaning and value ranges of each parameter against the schema. The text content is a developer-oriented capability explanation narration, with a slightly slowed speech rate and pitch consistent with the default, suitable for use as a product demo audio track, API documentation sample audio, or as a reference baseline for real-time announcement and intelligent customer service voice output. It should be noted that the Flash tier is priced lower than the Plus in the same family ($15 vs $20 per million characters), with an average WER/CER of 3.87, the lowest in the family, and a speaker similarity of 80.44, slightly lower than Plus; voice cloning is supported at the underlying level but not opened in the first version, nor does it support text-only voice design, so this piece only demonstrates the built-in timbre and natural language instruction control.