🎭 Qwen3-TTS with Emotion Tags
SpragAI/qwen3-tts-emotion-tags is a LoRA finetune of
Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice that adds inline
emotion control to the transcript: prefix your text with a tag such as
[Angry], [Sad] or [Happy] and the whole utterance is delivered in that
emotion. The speaker embeddings are left bit-identical to the base model, so
the nine preset voices sound exactly the same — only the delivery changes.
Tags: [Angry] [Sad] [Happy] [Fast] [Gentle] [Tired] [Fearful] [Disgusted] [Surprised]
The finetune was trained on English — other languages may work via the base model but were not in the training corpus.
| Text to synthesize | Emotion tag (prepended to the text) | Voice preset |
|---|
Author-generated reference audio for the showcase sentence, one per tag, plus the untagged neutral baseline (from the model card — Apache-2.0).
Model: SpragAI/qwen3-tts-emotion-tags · Base: Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice · Inference: qwen-tts · License: Apache-2.0