Instructions to use devtcpip/cutewallet-voice with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Pocket-TTS
How to use devtcpip/cutewallet-voice with Pocket-TTS:
from pocket_tts import TTSModel import scipy.io.wavfile tts_model = TTSModel.load_model("devtcpip/cutewallet-voice") voice_state = tts_model.get_state_for_audio_prompt( "hf://kyutai/tts-voices/alba-mackenna/casual.wav" ) audio = tts_model.generate_audio(voice_state, "Hello world, this is a test.") # Audio is a 1D torch tensor containing PCM data. scipy.io.wavfile.write("output.wav", tts_model.sample_rate, audio.numpy()) - Notebooks
- Google Colab
- Kaggle
CuteWallet voices (ONNX)
On-device voices for the CuteWallet app's helper characters. The Pocket TTS models are exported to ONNX (onnxruntime, CPU) and come with each helper's voice prompt. One folder per language.
| folder | source model | license |
|---|---|---|
tr/ |
wite-tech/pocket-tts-turkish, a student of Kyutai Pocket TTS | CC-BY-4.0 |
en/ |
kyutai/pocket-tts with voices from the Kyutai catalog | CC-BY-4.0 |
Each folder holds:
flow_lm_main_int8.onnx: the backbone, stateful, with an explicit KV cache.flow_lm_flow_int8.onnx: the flow head.mimi_decoder.onnx: the streaming codec decoder.text_conditioner.onnx: the token embedding table.model_config.json- the tokenizer as JSON
voices.json<helper>.prompt.f32.bin: raw float32 voice prompts.
Credits: Kyutai (Pocket TTS) and wite-tech (Turkish fine-tune). Use of these models must follow the source models' terms: no voice impersonation or cloning without consent, and no deceptive or harmful content. The helper voices are synthetic designs; they belong to no real person.
- Downloads last month
- -