Qwen3-TTS-Tokenizer-12Hz-48kHz: 48kHz Output Decoder Model

[takuma104]

Hello, I've released Qwen3-TTS-Tokenizer-12Hz-48kHz, a fine-tuned variant of Qwen/Qwen3-TTS-Tokenizer-12Hz that outputs 48 kHz audio by appending one additional DecoderBlock (upsample rate ×2) to the decoder stack.

Since only the decoder is fine-tuned, the codec remains fully compatible with the original tokenizer: you can drop this model in as a replacement decoder for Qwen3-TTS to obtain higher-quality audio output without any changes to the encoder or the TTS model itself.

Training Code

https://github.com/takuma104/Qwen3-TTS-Experimental/blob/decoder_block_48k/finetuning/decoder_block_48k/train_gan.sh

Mel spectrogram comparison: