[ 01 ] // quickstart
[ 02 ] // readme
qwen3-tts
Qwen3-TTS: Qwen's open speech model. Preset voices, 10 languages, tone control from a text instruction, and a variant that clones a voice from 3 seconds of audio.
Get it
arcflare gen setup qwen3-tts --torch cudaarcflare gen tts "Hello from ArcFlare" -m qwen3-ttsIt writes a .wav to the current folder (-o to choose). The model gets its own Python environment, so it can't break another one. The first run downloads the weights. On an AMD GPU under Windows, use --torch-from <python> instead of --torch cuda. See Speech.
Variants
qwen3-tts: 1.7B CustomVoice. Preset voices, plus--instructfor tone and emotionqwen3-tts-0.6b: the same preset voices, less memoryqwen3-tts-clone: 1.7B Base. Clones a voice from about 3 seconds of audio
arcflare gen tts "Hello there" -m qwen3-tts --instruct "warm, a little excited"arcflare gen tts "Hello there" -m qwen3-tts-clone --ref me.wav --ref-text "what I say in the clip"Languages: Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish and Italian. Pass --lang en or let it detect.
Measured on an AMD Radeon 8060S (ROCm), an integrated GPU: the 0.6B model generates at roughly 8–9 seconds per second of audio, even with the model already loaded. A dedicated NVIDIA card is much faster.