[Bug] wav file seems to contain Chinese words at the end

[dcolley]

Describe the bug

I am trying to produce audio that switches between languages.
tts seems to split the paragraph at the correct point, however the audio output contains (what appears to be) Chinese words after the French sentence.

To Reproduce

tts --text "Translate this to french, How are you? (Comment ça va?)" --model_name "tts_models/multilingual/multi-dataset/xtts_v2" --speaker_idx "Annmarie Nele" --language_idx en --out_path speech1.wav
tts_models/multilingual/multi-dataset/xtts_v2 is already downloaded.
Using model: xtts
Text: Translate this to french, How are you? (Comment ça va?)
Text split into sentences.
Input: ['Translate this to french, How are you?', '(Comment ça va?)']
Processing time: 4.410
Real-time factor: 0.462
Saved TTS output to speech1.wav

Expected behavior

No response

Logs

Environment

{
    "CUDA": {
        "GPU": [],
        "available": false,
        "version": null
    },
    "Packages": {
        "PyTorch_debug": false,
        "PyTorch_version": "2.6.0",
        "TTS": "0.26.0",
        "numpy": "1.26.4"
    },
    "System": {
        "OS": "Darwin",
        "architecture": [
            "64bit",
            ""
        ],
        "processor": "arm",
        "python": "3.11.11",
        "version": "Darwin Kernel Version 24.3.0: Thu Jan  2 20:24:22 PST 2025; root:xnu-11215.81.4~3/RELEASE_ARM64_T6041"
    }
}

Additional context

No response

[dcolley]

speech1.wav.zip

[eginhard]

XTTS wasn't trained on code-switched data, so the behaviour in that case is undefined and issues can be expected. You could dive into the XTTS code though and insert additional language tokens at the right points.