Christina TTS 1.5 — Q4 GGUF
4-bit (Q4_0) GGUF build of
Christina TTS 1.5 for
CPU-only inference ,no Python, no PyTorch, no GPU. Runs with
qwen3-tts.cpp at roughly real-time speed on a modern desktop CPU.
About the Voice
Important disclaimer: The voice actress in this model is not Asami Imai or any other official voice actor. This is a fan recreation using synthetic and real data. The voice actress who lent her voice wishes to remain private.
Files
| File | Purpose |
|---|
qwen3-tts-0.6b-f16.gguf | Talker model, Q4_0 quantized (~1.0 GB) |
qwen3-tts-tokenizer-f16.gguf | Speech tokenizer / decoder (F16) |
christina*.spk | Precomputed speaker embeddings, one per voice |
Note: the .gguf filenames are kept as-is because qwen3-tts-cli expects those exact names in the model directory.
Voices
| Speaker file | Language |
|---|
christina.spk | English |
christina-jp.spk | Japanese |
christina-fr.spk | French |
christina-es.spk | Spanish |
christina-ru.spk | Russian |
christina-de.spk | German |
christina-zh.spk | Chinese |
Quickstart
1# Build the engine
2git clone https://github.com/predict-woo/qwen3-tts.cpp
3cd qwen3-tts.cpp && cmake -B build && cmake --build build -j
4
5# Download this repo (needs access), then synthesize:
6./build/qwen3-tts-cli \
7 -m /path/to/christina-TTS-1.5-q4 \
8 -s /path/to/christina-TTS-1.5-q4/christina-fr.spk \
9 -l fr \
10 -t "J'ai enfin réussi à terminer l'expérience." \
11 -o out.wav \
12 --temperature 0.7 --top-k 20 --top-p 0.85 --repetition-penalty 1.1
Languages: en, ja, fr, es, ru, de, zh (pass with -l, matching the chosen .spk).
For GPU inference or the Python/Transformers ecosystem, use the full-precision model instead:
Loke-60000/christina-TTS-1.5.
License & Usage
This model is free to download and use. However, no training data will be shared.
If you use this model in any project, please provide credit.
