cantonese-0.6b (v1.2)

female voice with adapter

ggk -m cantonese-0.6b-v1.2-q8_0.gguf --adapter pig_cantonese_0.6b_adapter-f16.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, female voice, speaking slowly." -t "我覺得自己講到一嚿嚿，但佢應該明我意思啩。" -o out.wav

male voice with adapter

ggk -m cantonese-0.6b-v1.2-q8_0.gguf --adapter pig_cantonese_0.6b_adapter-f16.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, deep male voice, speaking slowly." -t "我覺得自己講到一嚿嚿，但佢應該明我意思啩。" -o out.wav

chicken theory (with adapter)

ggk -m cantonese-0.6b-v1.2-q8_0.gguf --adapter pig_cantonese_0.6b_adapter-f16.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, low female voice, speaking slowly." -t "牛暫時養唔起，豬同羊都養唔起，咁先養雞，雞大咗生好多蛋，蛋又生好多雞，再賣咗啲雞換羊，羊大咗再換豬，豬大咗再換牛。" -o out.wav

ggk -m cantonese-0.6b-v1.2-q8_0.gguf --adapter pig_cantonese_0.6b_adapter-f16.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, deep male voice, speaking slowly." -t "牛暫時養唔起，豬同羊都養唔起，咁先養雞，雞大咗生好多蛋，蛋又生好多雞，再賣咗啲雞換羊，羊大咗再換豬，豬大咗再換牛。" -o out.wav

ggk -m cantonese-0.6b-v1.2-q8_0.gguf --adapter pig_cantonese_0.6b_adapter-f16.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, low female voice, speaking slowly." -t "牛暫時養唔起，豬同羊都養唔起，咁先養雞，雞大咗生好多蛋，蛋又生好多雞，再賣咗啲雞換羊，羊大咗再換豬，豬大咗再換牛。" -o out.mp3

chicken theory (without adapter)

ggk -m cantonese-0.6b-v1.2-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, low male voice, speaking slowly." -t "牛暫時養唔起，豬同羊都養唔起，咁先養雞，雞大咗生好多蛋，蛋又生好多雞，再賣咗啲雞換羊，羊大咗再換豬，豬大咗再換牛" -o out.wav

developer note: v1.2 is able to produce more natural and fluent speech, good for voice clone but has issues on voice design, while v1.3 below performs better on voice design but the output might not as natural as v1.2, hence, an adapter was trained with v1.3 dataset (since v1.2 was trained with a totally different dataset), hopefully, v1.2 + adapter will take the benefit from both sides, able to give a more natural, fluent speech and a better voice control

cantonese-0.6b (v1.3)

ggk -m cantonese-0.6b-v1.3-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, low female voice, speaking slowly." -t "牛暫時養唔起，豬同羊都養唔起，咁先養雞，雞大咗生好多蛋，蛋又生好多雞，再賣咗啲雞換羊，羊大咗再換豬，豬大咗再換牛。" -o out.wav

ggk -m cantonese-0.6b-v1.3-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, low male voice, speaking slowly." -t "牛暫時養唔起，豬同羊都養唔起，咁先養雞，雞大咗生好多蛋，蛋又生好多雞，再賣咗啲雞換羊，羊大咗再換豬，豬大咗再換牛。" -o out.wav

cantonese-1.7b (v1.1)

voice design

ggk -m cantonese-1.7b-v1.1-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, low male voice, speaking slowly." -t "極少人會用呢個詞嚟表達，但我估佢應該明我意思。" -o out.wav

voice clone

ggk -m cantonese-1.7b-v1.1-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf -r out_ref.wav -t "極少人會講呢個詞，但佢應該明我意思啩。" -o out.wav

voice clone with reference text

ggk -m cantonese-1.7b-v1.1-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf -r out_ref.wav --reference-text "今日天氣好好，我哋出去行下街啦。" -t "極少人會用呢個詞嚟表達，但佢應該明我意思啩。" -o out.wav

resource

cantonese-1.7b-v1.1

cantonese-0.6b (v1)

female voice example

ggk -m cantonese-0.6b-v1a-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --speaker yue_f1 -t "今日天氣好好，我哋出去行下街啦。" -o out.wav

male voice example

ggk -m cantonese-0.6b-v1b-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --speaker yue_m1 -t "今日天氣好好，我哋出去行下街啦。" -o out.wav

cantonese-0.6b (v1.1)

voice clone

ggk -m cantonese-0.6b-v1.1-q8_0.gguf --codec-model qwen-tokenizer-12hz-q8_0.gguf -r out_ref.wav -t "極少人會講呢個詞，但佢應該明我意思啩。" -o out.wav

voice clone with reference text

ggk -m cantonese-0.6b-v1.1-q8_0.gguf --codec-model qwen-tokenizer-12hz-q8_0.gguf -r out_ref.wav --reference-text "今日天氣好好，我哋出去行下街啦。" -t "極少人會用呢個詞嚟表達，但佢應該明我意思啩。" -o out.wav

cantonese-1.7b (v1)

female voice 1 example

ggk -m cantonese-v1-nvfp4.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --speaker yue_f1 -t "今日天氣好好，我哋出去行下街啦。" -o out.wav

male voice 1 example

ggk -m cantonese-v1-nvfp4.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --speaker yue_m1 -t "今日天氣好好，我哋出去行下街啦。" -o out.wav

resource

cantonese-v1-dev

hk-cantonese (v0.0.5)

ggk -m hk-cantonese-v0.0.5-nvfp4.gguf --codec-model qwen-tokenizer-12hz-q8_0.gguf -t "今日天氣好好，我哋出去行下街啦。" -o out.wav

resource

hk-cantonese-cc0-qwen3-tts

english

ggk -m qwen3-tts-0.6b-base-q8_0.gguf --codec-model qwen-tokenizer-12hz-q8_0.gguf -t "Welcome to GGUF! The selected TTS model is now running." -o out.wav

voice clone (example)

ggk -m qwen3-tts-0.6b-base-q8_0.gguf --codec-model qwen-tokenizer-12hz-q8_0.gguf -r clone_ref_sample.wav -t "Hello from the cloned voice, this is a test." -o out_clone.wav

ggk -m qwen3-tts-0.6b-base-q8_0.gguf --codec-model qwen-tokenizer-12hz-q8_0.gguf -r clone_ref_sample.wav --reference-text "Hello from the cloned voice, this is a test." -t "The quick brown fox jumps over the lazy dog." -o out_clone_icl.wav

voice design (example)

ggk -m qwen3-tts-1.7b-voicedesign-q8_0.gguf --codec-model qwen-tokenizer-12hz-nvfp4.gguf --instruction "A warm, low female voice, speaking slowly." -t "Hello, this is a voice design test." -o out_voicedesign.wav

get tts tokenizer from gguf-connector