inclusionAI/Ming-3.0-tts-0.5B · Hugging Face

Source: https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B

Dense 0.5B multilingual speech synthesis with paralinguistic and free-form control.

GitHub|Blog & Demos|ModelScope

This repository containsMing-3.0-tts-0.5B, the dense Ming-3.0-tts checkpoint built on Qwen2.5-0.5B. The checkpoint includes the backbone, DiT head, audio tokenizer, speaker encoder and text tokenizer.

**Inference requires theMing-3.0-tts GitHub repository.**This Hugging Face repository is for checkpoint distribution and does not provide a standalone inference setup. Clone the GitHub repository and follow its README for the current dependencies, inference code and usage instructions. The CLI scripts and Python APIs referenced below are provided by the GitHub repository.

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#model-detailsModel Details

ItemValueArchitectureDense, Qwen2.5-0.5B backboneParameters0.5BLanguagesChinese, English, German, Spanish, French, Korean, JapaneseDialectsSichuanese, CantonesePython APIming\_tts\.MingTTSCLIinfer\.pySample rate44.1 kHz

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#capabilitiesCapabilities

• Zero-shot voice cloning from a short reference clip
• Voice Design from a natural-language speaker description
• Free-form control of delivery and speaking style
• 100+ built-in Chinese voices and five built-in English voices
• 14 inline tags for breath, cough, sigh, laughter, hesitation, surprise and question/confirmation sounds in Chinese or English
• Explicit emotion, speed, pitch and volume control
• Multilingual synthesis and Sichuanese/Cantonese dialect control

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#quick-startQuick Start

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#installationInstallation

git clone https://github.com/inclusionAI/Ming-3.0-tts.git
cd Ming-3.0-tts
pip install -r requirements.txt

Tested on NVIDIA H800-80GB and H20-96G with CUDA 12.4.flash\-attnis optional; inference falls back to PyTorch SDPA.

Numbers, dates, units and symbols are normalized with FST grammars. Installpynini, or pass\-\-no\_text\_normalizationto skip text normalization:

conda install -c conda-forge pynini==2.1.6

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#downloadDownload

pip install "huggingface_hub[cli]"
hf download inclusionAI/Ming-3.0-tts-0.5B --local-dir Ming-3.0-tts-0.5b

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#cli-inferenceCLI Inference

python infer.py \
  --model_path inclusionAI/Ming-3.0-tts-0.5B \
  --task TTS \
  --language_id Chinese \
  --text "傍晚的风从山口吹过来，远处的云慢慢染成了金色。" \
  --prompt_wav_path data/wavs/10002287-00000094.wav \
  --prompt_text "在此奉劝大家别乱打美白针。" \
  --use_spk_emb \
  --output output/tts.wav

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#python-apiPython API

from ming_tts import MingTTS

model = MingTTS("inclusionAI/Ming-3.0-tts-0.5B")
model.synthesize(
    task="TTS",
    language_id="Chinese",             # English for English text
    text="傍晚的风从山口吹过来，远处的云慢慢染成了金色。",
    prompt_wav_path="data/wavs/10002287-00000094.wav",
    prompt_text="在此奉劝大家别乱打美白针。",
    use_spk_emb=True,
    output_path="output/tts.wav",
)

Supported tasks areTTS,FREEFORM,VOICE\_DESIGN,MULTILINGUAL,BASIC,EMOTION,DIALECTandIP. See thetask cookbookfor controls and examples.

MULTILINGUALrequireslanguage\_idfor German, Spanish, French, Korean or Japanese;DIALECTtakes no language ID. Freeform uses its own task and cannot be combined with voice design, a built-in voice, emotion, dialect or attribute controls. With a reference clip it follows that voice; without a reference, the instruction can choose the voice.

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#verify-the-checkpointVerify the Checkpoint

python tools/preflight.py --model_path /path/to/Ming-3.0-tts-0.5b

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#performancePerformance

Benchmark comparisons from theproject README.

Boldandunderlinemark the best and second-best values among the listed open-source models; ties share a mark. Seed-TTS and NVbench-GT are unranked reference rows.

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#zero-shot-tts-seed-tts-testsetZero-shot TTS (Seed-TTS testset)

Modelseed-tts-eval-zhseed-tts-eval-enWER ↓SIM ↑WER ↓SIM ↑Seed-TTS1.110%0.7962.240%0.762GLM-TTS1.030%0.7612.230%0.672Qwen3-TTS1.070%—1.390%—Qwen3-TTS-12Hz-0.6B-Base0.920%—1.320%—Qwen3-TTS-12Hz-1.7B-Base0.770%—1.240%—CosyVoice3-0.5B1.160%0.7802.020%0.718CosyVoice3-1.5B0.710%0.7751.450%0.695F5-TTS1.560%0.7411.830%0.647Ming-3.0-tts-0.5b0.800%0.7581.584%0.691Ming-3.0-tts-tiny0.820%0.7811.666%0.728### https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#multilingual-minimax-test-set-selected-languagesMultilingual (MiniMax test set, selected languages)

Evaluated on the German, Spanish, French, Korean, and Japanese subsets ofMiniMaxAI/TTS-Multilingual-Test-Set.

ModelGermanSpanishFrenchKoreanJapaneseWER ↓SIM ↑WER ↓SIM ↑WER ↓SIM ↑CER ↓SIM ↑CER ↓SIM ↑dots.tts0.452%0.7982.593%0.83919.878%0.7894.837%0.8535.134%0.838VoxCPM20.963%0.7901.324%0.836**3.380%0.7613.547%0.8394.358%0.825Fish Audio S21.045%0.7141.216%0.7483.636%0.6981.675%0.7613.427%0.755FireRedTTS30.348%0.8251.086%0.8684.366%0.807****1.644%0.8643.418%0.825Ming-3.0-tts-0.5b0.365%0.7531.099%0.7634.264%0.6751.728%0.7933.765%0.765Ming-3.0-tts-tiny****0.325%0.7910.971%**0.7973.470%0.7336.000%0.7973.427%0.780Each open-source model was evaluated using its recommended inference settings under a unified ASR-based WER/CER evaluation protocol.

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#paralinguistic-control-nvbenchParalinguistic Control (NVBench)

ModelSingleMultiChineseEnglishChineseEnglishCERPCEROCERCERPCEROCERCERPCEROCERCERPCEROCERNVbench-GT3.869.384.076.738.316.903.7923.715.077.2621.418.62Orpheus-TTS11.3688.7713.919.0371.9210.6319.8384.8524.388.6871.4611.89SMIIP-NV-CV28.8075.6411.3417.9256.8019.4710.6677.2014.7920.9354.4923.87Emilia-NV-CV25.0540.006.6412.5055.3013.215.5448.748.0911.7160.2814.63CosyVoice33.8557.695.867.8762.759.064.7561.948.266.3957.8410.69NV-FlexiVoice6.9831.088.1511.8850.4313.218.2039.3710.399.6051.3213.76NV-CV33.8027.694.908.3346.139.443.4430.044.846.7047.1310.10Ming-3.0-tts-0.5b****2.7022.803.207.6037.008.903.5036.608.107.9047.3010.40Ming-3.0-tts-tiny2.8017.233.205.8632.957.054.4932.885.815.1246.137.68### https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#dialect-controlDialect Control

ModelWSC-Eval-TTS-easyWSC-Eval-TTS-hardWSYue-TTS-eval-BaseWSYue-TTS-eval-CoverageCER ↓SIM ↑ACC ↑CER ↓SIM ↑ACC ↑CER ↓SIM ↑ACC ↑CER ↓SIM ↑ACC ↑Qwen3-TTS4.13%——7.35%————————CosyVoice3-0.5B3.17%0.69676.05%4.07%0.72381.41%8.36%0.61191.70%8.95%0.658**95.80%**CosyVoice3-1.5B + DiffRO-EMO7.14%0.703—9.06%0.601—14.38%0.812—13.74%0.826—CosyVoice2-WSC4.28%0.728—8.87%0.626———————CosyVoice2-WSC-SFT4.08%0.788—7.22%0.680———————CosyVoice2-Yue——————10.33%0.821—9.49%0.834—Sparks-TTS10.83%0.677—12.52%0.545—27.79%0.762—24.25%0.781—**Ming-3.0-tts-0.5b****2.27%**0.68980.68%3.35%0.69187.64%**6.15%0.58698.00%****8.15%0.62295.27%Ming-3.0-tts-tiny2.53%0.72186.09%****3.11%0.73888.94%**7.40%0.60592.90%8.67%0.65395.16%### https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#emotion-control-cv3Emotion Control (CV3)

ModelAverageText-RelatedText-UnrelatedHappySadAngryHappySadAngryCosyVoice3-0.5B66.33%92%70%72%64%42%58%CosyVoice3-1.5B63.00%86%64%72%64%44%48%CosyVoice3-1.5B + DiffRO-EMO**77.67%****98%**68%84%**98%**50%68%F5-TTS64.67%92%52%72%80%28%64%Sparks-TTS55.33%80%56%50%50%**60%**36%Ming-3.0-tts-0.5b72.33%96%76%**94%**52%36%80%****Ming-3.0-tts-tiny74.00%**98%****82%**90%60%36%78%### https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#emotion-control-cv3n-hard-emotion-controlEmotion Control (CV3N, hard emotion control)

ModelAverageText-RelatedText-UnrelatedHappySadAngryHappySadAngryCosyVoice3-0.5B**40.00%**68%30%**78%**14%4%46%****Ming-3.0-tts-0.5b37.33%**70%**36%70%14%4%30%Ming-3.0-tts-tiny30.67%60%**40%**44%**18%****10%**12%### https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#voice-design-instructttseval-zhVoice Design (InstructTTSEval-ZH)

ModelAverageAPSDSDRPQwen3-TTS**77.13%****85.20%****81.10%****65.10%**Mimo-Audio-7B-Instruct70.50%75.70%74.30%61.50%VoiceSculptor67.30%75.70%64.70%61.50%VoxInstruct47.47%47.50%52.30%42.60%Ming-3.0-tts-0.5b72.29%78.54%77.28%61.06%Ming-3.0-tts-tiny75.37%85.10%79.90%61.10%### https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#basic-attribute-controlBasic Attribute Control

ModelSpeedVolumePitchAverageWER ↓SIM ↑CosyVoice3-0.5B100.00%****97.67%65.33%87.67%1.210%0.580Ming-3.0-tts-0.5b99.00%95.00%86.00%****93.30%0.317%0.712Ming-3.0-tts-tiny93.00%93.00%78.70%88.23%**0.274%**0.742## https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#limitationsLimitations

• Speaker similarity is weaker than content accuracy; French SIM is the clearest multilingual gap.
• NVBench Multi trails Single; stability with several paralinguistic events in one sentence is still improving.
• Emotion control is harder when the words carry no emotional signal.
• Pitch is less accurate than speed and volume, and open-ended Voice Design does not yet match the strongest evaluated system.

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#usage-disclaimerUsage Disclaimer

• Voice cloning and built-in character voices are provided for research and legitimate creative use. Do not use them to impersonate a real person without consent, to deceive, or for any illegal purpose.
• Label synthetic speech as synthetic where it could be mistaken for a real recording.
• The authors accept no liability for misuse. If you find abuse of this project, please report it to us.

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#acknowledgementsAcknowledgements

• Qwen2.5for the backbone
• CAM++for speaker embedding extraction
• Ming-omni-tts, the previous generation, whose audio tokenizer this release builds on

https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B#licenseLicense

Released under theMIT license.