{
  "name": "LLM Configurator voice model catalog",
  "url": "https://llmconfigurator.com/en/voice",
  "licence": "CC-BY-4.0",
  "updated": "2026-10-25",
  "notes": [
    "licence.commercialClass: C = weights allow commercial use; C* = with conditions; NC = non-commercial; ? = not stated on the model card.",
    "memory.publishedGB is what the maker or a named source publishes. It is not measured and not a fit verdict.",
    "leaderboard values carry their board and date; never compare two boards."
  ],
  "models": [
    {
      "id": "whisper-large-v3-turbo",
      "name": "Whisper large-v3-turbo",
      "maker": "OpenAI",
      "category": "stt",
      "licence": {
        "weights": "MIT",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [
        "cpu"
      ],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": true,
      "languages": 90,
      "runtimes": [
        "whisper.cpp",
        "faster-whisper",
        "WhisperKit",
        "Transformers",
        "FunASR"
      ],
      "repo": "https://huggingface.co/openai/whisper-large-v3-turbo",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 7.64,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "10 Hugging Face Models Worth Knowing in 2026",
          "url": "https://www.kdnuggets.com/10-hugging-face-models-worth-knowing-in-2026",
          "checked": "2026-10-25"
        },
        {
          "title": "FunASR on PyPI (v1.3.16 changelog, model table)",
          "url": "https://pypi.org/project/funasr/1.3.16/",
          "checked": "2026-10-25"
        },
        {
          "title": "openai/whisper-large-v3-turbo model card",
          "url": "https://huggingface.co/openai/whisper-large-v3-turbo",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "qwen3-asr-1-7b",
      "name": "Qwen3-ASR 1.7B",
      "maker": "Alibaba Qwen",
      "category": "stt",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "16-24gb",
      "alsoTier": [
        "8gb"
      ],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": null,
      "languages": 52,
      "runtimes": [
        "Transformers",
        "vLLM",
        "FunASR",
        "llama.cpp (reported, verify build)"
      ],
      "repo": "https://huggingface.co/Qwen/Qwen3-ASR-1.7B",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 5.4,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best open-source speech-to-text models in 2026",
          "url": "https://www.gladia.io/blog/best-open-source-speech-to-text-models",
          "checked": "2026-10-25"
        },
        {
          "title": "FunASR on PyPI (v1.3.16 changelog, model table)",
          "url": "https://pypi.org/project/funasr/1.3.16/",
          "checked": "2026-10-25"
        },
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "Underdog-Bark-3.0-Pro model card (reprints Open ASR Leaderboard rows as of 2026-10-09)",
          "url": "https://huggingface.co/ConwayResearch/Underdog-Bark-3.0-Pro",
          "checked": "2026-10-25"
        },
        {
          "title": "Qwen/Qwen3-ASR-1.7B model card",
          "url": "https://huggingface.co/Qwen/Qwen3-ASR-1.7B",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "qwen3-asr-0-6b",
      "name": "Qwen3-ASR 0.6B",
      "maker": "Alibaba Qwen",
      "category": "stt",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [
        "cpu"
      ],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": null,
      "languages": 52,
      "runtimes": [
        "Transformers",
        "vLLM",
        "FunASR"
      ],
      "repo": "https://huggingface.co/Qwen/Qwen3-ASR-0.6B",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 6.24,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best open-source speech-to-text models in 2026",
          "url": "https://www.gladia.io/blog/best-open-source-speech-to-text-models",
          "checked": "2026-10-25"
        },
        {
          "title": "Underdog-Bark-3.0-Pro model card (reprints Open ASR Leaderboard rows as of 2026-10-09)",
          "url": "https://huggingface.co/ConwayResearch/Underdog-Bark-3.0-Pro",
          "checked": "2026-10-25"
        },
        {
          "title": "Qwen/Qwen3-ASR-0.6B model card",
          "url": "https://huggingface.co/Qwen/Qwen3-ASR-0.6B",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "cohere-transcribe-03-2026",
      "name": "Cohere Transcribe",
      "maker": "Cohere Labs",
      "category": "stt",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "16-24gb",
      "alsoTier": [
        "8gb"
      ],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": null,
      "languages": 14,
      "runtimes": [
        "Transformers",
        "vLLM"
      ],
      "repo": "https://huggingface.co/CohereLabs/cohere-transcribe-03-2026",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 5.77,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "CohereLabs/cohere-transcribe-03-2026 model card",
          "url": "https://huggingface.co/CohereLabs/cohere-transcribe-03-2026",
          "checked": "2026-10-25"
        },
        {
          "title": "Introducing Cohere-transcribe",
          "url": "https://huggingface.co/blog/CohereLabs/cohere-transcribe-03-2026-release",
          "checked": "2026-10-25"
        },
        {
          "title": "Open ASR Leaderboard: Towards Reproducible and Transparent Multilingual and Long-Form Speech Recognition Evaluation",
          "url": "https://arxiv.org/pdf/2510.06961",
          "checked": "2026-10-25"
        },
        {
          "title": "CohereLabs/cohere-transcribe-03-2026 model card",
          "url": "https://huggingface.co/CohereLabs/cohere-transcribe-03-2026",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "canary-qwen-2-5b",
      "name": "Canary-Qwen 2.5B",
      "maker": "NVIDIA",
      "category": "stt",
      "licence": {
        "weights": "CC-BY-4.0",
        "commercialClass": "C",
        "condition": "Attribution required.",
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "16-24gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": null,
      "languages": 1,
      "runtimes": [
        "NeMo"
      ],
      "repo": "https://huggingface.co/nvidia/canary-qwen-2.5b",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 5.94,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best open-source speech-to-text models in 2026",
          "url": "https://www.gladia.io/blog/best-open-source-speech-to-text-models",
          "checked": "2026-10-25"
        },
        {
          "title": "Best open source speech-to-text (STT) model in 2026",
          "url": "https://northflank.com/blog/best-open-source-speech-to-text-stt-model-in-2026-benchmarks",
          "checked": "2026-10-25"
        },
        {
          "title": "Open ASR Leaderboard: Towards Reproducible and Transparent Multilingual and Long-Form Speech Recognition Evaluation",
          "url": "https://arxiv.org/pdf/2510.06961",
          "checked": "2026-10-25"
        },
        {
          "title": "nvidia/canary-qwen-2.5b model card",
          "url": "https://huggingface.co/nvidia/canary-qwen-2.5b",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "parakeet-tdt-0-6b-v3",
      "name": "Parakeet TDT v3",
      "maker": "NVIDIA",
      "category": "stt",
      "licence": {
        "weights": "CC-BY-4.0",
        "commercialClass": "C",
        "condition": "Attribution required.",
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [],
      "memory": {
        "publishedGB": 2,
        "basis": "At least 2 GB RAM to load (card).",
        "publishedBy": "NVIDIA (model card)",
        "singleSource": false
      },
      "voiceCloning": null,
      "cpuOK": null,
      "languages": 25,
      "runtimes": [
        "NeMo",
        "Transformers"
      ],
      "repo": "https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 6.18,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "parakeet-tdt-0.6b-v3 model card",
          "url": "https://build.nvidia.com/nvidia/parakeet-tdt-0_6b/modelcard",
          "checked": "2026-10-25"
        },
        {
          "title": "Open ASR Leaderboard: Towards Reproducible and Transparent Multilingual and Long-Form Speech Recognition Evaluation",
          "url": "https://arxiv.org/pdf/2510.06961",
          "checked": "2026-10-25"
        },
        {
          "title": "nvidia/parakeet-tdt-0.6b-v3 model card",
          "url": "https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "moonshine-v2",
      "name": "Moonshine v2",
      "maker": "Moonshine AI",
      "category": "stt",
      "licence": {
        "weights": "MIT",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "cpu",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": true,
      "languages": null,
      "runtimes": [
        "ONNX (Moonshine Voice)",
        "Transformers"
      ],
      "repo": "https://huggingface.co/moonshine-ai/moonshine-streaming-medium",
      "leaderboards": [],
      "sources": [
        {
          "title": "Moonshine v2: Ergodic Streaming Encoder ASR for Latency-Critical Speech",
          "url": "https://arxiv.org/pdf/2602.12241",
          "checked": "2026-10-25"
        },
        {
          "title": "Real-time Speech Recognition on SL2610 (Moonshine V2 on an NPU)",
          "url": "https://developer.synaptics.com/blog/realtime-asr",
          "checked": "2026-10-25"
        },
        {
          "title": "moonshine-ai/moonshine-streaming-medium model card",
          "url": "https://huggingface.co/moonshine-ai/moonshine-streaming-medium",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "fun-asr-nano",
      "name": "Fun-ASR-Nano",
      "maker": "FunAudioLLM (Alibaba Tongyi)",
      "category": "stt",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "cpu",
      "alsoTier": [
        "8gb"
      ],
      "memory": {
        "publishedGB": 0.95,
        "basis": "encoder f16 470 MB + Qwen3-0.6B q4_K_M 484 MB GGUF",
        "publishedBy": "nile2000 on Hugging Face",
        "singleSource": true
      },
      "voiceCloning": null,
      "cpuOK": true,
      "languages": 31,
      "runtimes": [
        "FunASR llama.cpp runtime (single binary, no Python)",
        "vLLM"
      ],
      "repo": "https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 7.24,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "FunASR on PyPI (v1.3.16 changelog, model table)",
          "url": "https://pypi.org/project/funasr/1.3.16/",
          "checked": "2026-10-25"
        },
        {
          "title": "Fun-ASR-Nano GGUF (FunASR llama.cpp runtime)",
          "url": "https://huggingface.co/nile2000/Fun-ASR-Nano-GGUF",
          "checked": "2026-10-25"
        },
        {
          "title": "FunAudioLLM/Fun-ASR-Nano-2512 model card",
          "url": "https://huggingface.co/FunAudioLLM/Fun-ASR-Nano-2512",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "nemotronlabs-voicechat-11b",
      "name": "VoiceChat 11B",
      "maker": "NVIDIA",
      "category": "s2s",
      "licence": {
        "weights": "OpenMDW-1.1",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "server",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": false,
      "languages": null,
      "runtimes": [
        "vLLM",
        "NeMo Speech"
      ],
      "repo": "https://huggingface.co/nvidia/NVIDIA-NemotronLabs-VoiceChat-11B",
      "leaderboards": [],
      "sources": [
        {
          "title": "nvidia/NVIDIA-NemotronLabs-VoiceChat-11B model card",
          "url": "https://huggingface.co/nvidia/NVIDIA-NemotronLabs-VoiceChat-11B",
          "checked": "2026-10-25"
        },
        {
          "title": "NVIDIA Releases NemotronLabs VoiceChat, an Open Full-Duplex Voice Model",
          "url": "https://rits.shanghai.nyu.edu/ai/nvidia-releases-nemotronlabs-voicechat-an-open-full-duplex-voice-model",
          "checked": "2026-10-25"
        },
        {
          "title": "NVIDIA Opens NemotronLabs VoiceChat 11B With Tool Calling",
          "url": "https://aiweekly.co/alerts/nvidia-opens-nemotronlabs-voicechat-11b-with-tool-calling",
          "checked": "2026-10-25"
        },
        {
          "title": "nvidia/NVIDIA-NemotronLabs-VoiceChat-11B model card",
          "url": "https://huggingface.co/nvidia/NVIDIA-NemotronLabs-VoiceChat-11B",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "minicpm-o-4-5",
      "name": "MiniCPM-o 4.5",
      "maker": "OpenBMB / ModelBest",
      "category": "s2s",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "16-24gb",
      "alsoTier": [
        "8gb"
      ],
      "memory": {
        "publishedGB": 12,
        "basis": "Full-duplex through llama.cpp-omni: an NVIDIA GPU with at least 12 GB, or an Apple M4 Max with at least 24 GB (card). llama.cpp-omni lists the full omni stack at Q4_K_M as ~8 GB model / ~9 GB peak on NVIDIA.",
        "publishedBy": "OpenBMB / ModelBest (model card)",
        "singleSource": false
      },
      "voiceCloning": null,
      "cpuOK": true,
      "languages": null,
      "runtimes": [
        "llama.cpp-omni",
        "Ollama (per card)",
        "vLLM",
        "SGLang",
        "official Docker image for Mac"
      ],
      "repo": "https://huggingface.co/openbmb/MiniCPM-o-4_5",
      "leaderboards": [],
      "sources": [
        {
          "title": "openbmb/MiniCPM-o-4_5 model card",
          "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5",
          "checked": "2026-10-25"
        },
        {
          "title": "llama.cpp-omni README (memory table)",
          "url": "https://github.com/tc-mb/llama.cpp-omni",
          "checked": "2026-10-25"
        },
        {
          "title": "openbmb/MiniCPM-o-4_5 model card",
          "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "moshi",
      "name": "Moshi",
      "maker": "Kyutai",
      "category": "s2s",
      "licence": {
        "weights": "CC-BY-4.0",
        "commercialClass": "C",
        "condition": "Attribution required; the card forbids impersonation.",
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "16-24gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": false,
      "cpuOK": false,
      "languages": null,
      "runtimes": [
        "official demo server"
      ],
      "repo": "https://huggingface.co/kyutai/moshiko-pytorch-bf16",
      "leaderboards": [],
      "sources": [
        {
          "title": "Full-Duplex-Bench v1.5",
          "url": "https://arxiv.org/pdf/2507.23159",
          "checked": "2026-10-25"
        },
        {
          "title": "NVIDIA Releases NemotronLabs VoiceChat, an Open Full-Duplex Voice Model",
          "url": "https://rits.shanghai.nyu.edu/ai/nvidia-releases-nemotronlabs-voicechat-an-open-full-duplex-voice-model",
          "checked": "2026-10-25"
        },
        {
          "title": "kyutai/moshiko-pytorch-bf16 model card",
          "url": "https://huggingface.co/kyutai/moshiko-pytorch-bf16",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "kokoro-82m",
      "name": "Kokoro 82M",
      "maker": "hexgrad",
      "category": "tts",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "cpu",
      "alsoTier": [
        "8gb"
      ],
      "memory": {
        "publishedGB": 2,
        "basis": "under 2 GB VRAM, or ~4 GB RAM CPU-only via ONNX",
        "publishedBy": "Pinggy Blog",
        "singleSource": true
      },
      "voiceCloning": false,
      "cpuOK": true,
      "languages": 8,
      "runtimes": [
        "kokoro-onnx",
        "Kokoro-FastAPI",
        "LocalAI"
      ],
      "repo": "https://huggingface.co/hexgrad/Kokoro-82M",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1066,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "9 Best Open Source Text to Speech Models in 2026",
          "url": "https://www.bland.ai/blog/best-open-source-text-to-speech-model",
          "checked": "2026-10-25"
        },
        {
          "title": "hexgrad/Kokoro-82M model card",
          "url": "https://huggingface.co/hexgrad/Kokoro-82M",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "piper",
      "name": "Piper",
      "maker": "Open Home Foundation (OHF-Voice)",
      "category": "tts",
      "licence": {
        "weights": "per voice (code GPL-3.0)",
        "commercialClass": "?",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": null,
      "tier": "cpu",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": false,
      "cpuOK": true,
      "languages": null,
      "runtimes": [
        "piper CLI",
        "Home Assistant Wyoming",
        "LocalAI"
      ],
      "repo": null,
      "leaderboards": [],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "voiceai: awesome list for voice agents",
          "url": "https://github.com/mahimairaja/voiceai",
          "checked": "2026-10-25"
        },
        {
          "title": "rhasspy/piper-voices model card",
          "url": "https://huggingface.co/rhasspy/piper-voices",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "pocket-tts",
      "name": "Pocket TTS",
      "maker": "Kyutai",
      "category": "tts",
      "licence": {
        "weights": "CC-BY-4.0",
        "commercialClass": "C",
        "condition": "Attribution required.",
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "cpu",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": true,
      "languages": 6,
      "runtimes": [
        "pocket-tts (pip/uvx)",
        "ONNX INT8 export",
        "browser (WASM)"
      ],
      "repo": "https://huggingface.co/kyutai/pocket-tts",
      "leaderboards": [],
      "sources": [
        {
          "title": "Pocket TTS review: Kyutai's 100M CPU voice cloner",
          "url": "https://dev.to/andrew-ooo/pocket-tts-review-kyutais-100m-cpu-voice-cloner-2026-40lb",
          "checked": "2026-10-25"
        },
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "Pocket TTS 100M ONNX INT8 (licence notes)",
          "url": "https://huggingface.co/soniqo/Pocket-TTS-100M-ONNX-INT8",
          "checked": "2026-10-25"
        },
        {
          "title": "kyutai/pocket-tts model card",
          "url": "https://huggingface.co/kyutai/pocket-tts",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "qwen3-tts",
      "name": "Qwen3-TTS",
      "maker": "Alibaba Qwen",
      "category": "tts",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [],
      "memory": {
        "publishedGB": 4.54,
        "basis": "1.7B download 4.54 GB; 0.6B 2.52 GB",
        "publishedBy": "Pinggy Blog",
        "singleSource": true
      },
      "voiceCloning": true,
      "cpuOK": null,
      "languages": 10,
      "runtimes": [
        "vLLM-Omni",
        "LocalAI",
        "Transformers"
      ],
      "repo": "https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-Base",
      "leaderboards": [],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "free-voice-clone (TTS/ASR model list with release dates)",
          "url": "https://github.com/0xSojalSec/free-voice-clone",
          "checked": "2026-10-25"
        },
        {
          "title": "Qwen/Qwen3-TTS-12Hz-1.7B-Base model card",
          "url": "https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-Base",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "chatterbox",
      "name": "Chatterbox",
      "maker": "Resemble AI",
      "category": "tts",
      "licence": {
        "weights": "MIT",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": null,
      "languages": 23,
      "runtimes": [
        "pip chatterbox-tts",
        "LocalAI"
      ],
      "repo": "https://huggingface.co/ResembleAI/chatterbox",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1025,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "9 Best Open Source Text to Speech Models in 2026",
          "url": "https://www.bland.ai/blog/best-open-source-text-to-speech-model",
          "checked": "2026-10-25"
        },
        {
          "title": "ResembleAI/chatterbox model card",
          "url": "https://huggingface.co/ResembleAI/chatterbox",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "voxcpm2",
      "name": "VoxCPM2",
      "maker": "OpenBMB",
      "category": "tts",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "16-24gb",
      "alsoTier": [
        "cpu"
      ],
      "memory": {
        "publishedGB": 8,
        "basis": "~8 GB VRAM (card).",
        "publishedBy": "OpenBMB (model card)",
        "singleSource": false
      },
      "voiceCloning": true,
      "cpuOK": true,
      "languages": 30,
      "runtimes": [
        "llama.cpp-omni (GGUF on CPU/Metal/CUDA/Vulkan)",
        "vLLM-Omni",
        "PyTorch"
      ],
      "repo": "https://huggingface.co/openbmb/VoxCPM2",
      "leaderboards": [],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "openbmb/VoxCPM2 model card",
          "url": "https://huggingface.co/openbmb/VoxCPM2",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "step-audio-editx",
      "name": "Step Audio EditX",
      "maker": "StepFun",
      "category": "tts",
      "licence": {
        "weights": "not stated (code Apache-2.0)",
        "commercialClass": "?",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": null,
      "tier": "16-24gb",
      "alsoTier": [],
      "memory": {
        "publishedGB": 12,
        "basis": "12 GB GPU memory is \"a critical value\"; 16 GB is safer (card).",
        "publishedBy": "StepFun (model card)",
        "singleSource": false
      },
      "voiceCloning": true,
      "cpuOK": false,
      "languages": 6,
      "runtimes": [
        "vLLM"
      ],
      "repo": "https://huggingface.co/stepfun-ai/Step-Audio-EditX",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1100,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "stepfun-ai/Step-Audio-EditX model card",
          "url": "https://huggingface.co/stepfun-ai/Step-Audio-EditX",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "voxtral-tts",
      "name": "Voxtral TTS",
      "maker": "Mistral AI",
      "category": "tts",
      "licence": {
        "weights": "CC-BY-NC-4.0",
        "commercialClass": "NC",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": false,
      "tier": "16-24gb",
      "alsoTier": [],
      "memory": {
        "publishedGB": 16,
        "basis": "A single GPU with at least 16 GB (card: BF16 weights).",
        "publishedBy": "Mistral AI (model card)",
        "singleSource": false
      },
      "voiceCloning": true,
      "cpuOK": false,
      "languages": 9,
      "runtimes": [
        "vLLM-Omni"
      ],
      "repo": "https://huggingface.co/mistralai/Voxtral-4B-TTS-2603",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1091,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "NVIDIA Releases NemotronLabs VoiceChat, an Open Full-Duplex Voice Model",
          "url": "https://rits.shanghai.nyu.edu/ai/nvidia-releases-nemotronlabs-voicechat-an-open-full-duplex-voice-model",
          "checked": "2026-10-25"
        },
        {
          "title": "mistralai/Voxtral-4B-TTS-2603 model card",
          "url": "https://huggingface.co/mistralai/Voxtral-4B-TTS-2603",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "fish-audio-s2-pro",
      "name": "Fish S2 Pro",
      "maker": "Fish Audio",
      "category": "tts",
      "licence": {
        "weights": "Fish Audio Research License",
        "commercialClass": "NC",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": false,
      "tier": "server",
      "alsoTier": [
        "16-24gb"
      ],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": false,
      "languages": 80,
      "runtimes": [
        "SGLang-Omni",
        "vLLM-Omni"
      ],
      "repo": "https://huggingface.co/fishaudio/s2-pro",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1110,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "free-voice-clone (TTS/ASR model list with release dates)",
          "url": "https://github.com/0xSojalSec/free-voice-clone",
          "checked": "2026-10-25"
        },
        {
          "title": "fishaudio/s2-pro model card",
          "url": "https://huggingface.co/fishaudio/s2-pro",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "breeze-tts-2",
      "name": "Breeze TTS 2",
      "maker": "BreezeBlue (RESONIA)",
      "category": "tts",
      "licence": {
        "weights": "BreezeBlue Research and Non-Commercial License",
        "commercialClass": "NC",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": false,
      "tier": "16-24gb",
      "alsoTier": [],
      "memory": {
        "publishedGB": 12,
        "basis": "~7.7 GiB eager; 12 GB GPU minimum, 24 GB for the --fast-all path (card).",
        "publishedBy": "BreezeBlue (RESONIA) (model card)",
        "singleSource": false
      },
      "voiceCloning": true,
      "cpuOK": false,
      "languages": 2,
      "runtimes": [
        "PyTorch CLI",
        "Docker"
      ],
      "repo": "https://huggingface.co/BreezeBlue/Breeze-TTS-2",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1230,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "BreezeBlue/Breeze-TTS-2 model card",
          "url": "https://huggingface.co/BreezeBlue/Breeze-TTS-2",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "higgs-audio-v3-tts",
      "name": "Higgs Audio V3",
      "maker": "Boson AI",
      "category": "tts",
      "licence": {
        "weights": "Boson Higgs TTS 3 Research and Non-Commercial License",
        "commercialClass": "NC",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": false,
      "tier": "server",
      "alsoTier": [
        "16-24gb"
      ],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": false,
      "languages": 100,
      "runtimes": [
        "SGLang-Omni"
      ],
      "repo": "https://huggingface.co/bosonai/higgs-tts-3-4b",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1045,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "bosonai/higgs-tts-3-4b model card",
          "url": "https://huggingface.co/bosonai/higgs-tts-3-4b",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "magpie-multilingual-357m",
      "name": "Magpie 357M",
      "maker": "NVIDIA",
      "category": "tts",
      "licence": {
        "weights": "NVIDIA Open Model License",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": false,
      "cpuOK": null,
      "languages": 12,
      "runtimes": [
        "NeMo"
      ],
      "repo": "https://huggingface.co/nvidia/magpie_tts_multilingual_357m",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1065,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "nvidia/magpie_tts_multilingual_357m model card",
          "url": "https://huggingface.co/nvidia/magpie_tts_multilingual_357m",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "omnivoice",
      "name": "OmniVoice",
      "maker": "k2-fsa",
      "category": "tts",
      "licence": {
        "weights": "CC-BY-NC",
        "commercialClass": "NC",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": false,
      "tier": "8gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": null,
      "languages": 600,
      "runtimes": [
        "vLLM-Omni"
      ],
      "repo": "https://huggingface.co/k2-fsa/OmniVoice",
      "leaderboards": [],
      "sources": [
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "k2-fsa/OmniVoice model card",
          "url": "https://huggingface.co/k2-fsa/OmniVoice",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "cosyvoice3-0-5b",
      "name": "Fun-CosyVoice3 0.5B",
      "maker": "FunAudioLLM",
      "category": "tts",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": null,
      "languages": null,
      "runtimes": [
        "vLLM-Omni"
      ],
      "repo": "https://huggingface.co/FunAudioLLM/Fun-CosyVoice3-0.5B-2512",
      "leaderboards": [],
      "sources": [
        {
          "title": "Open TTS Leaderboard: Scalable Evaluation for Multilingual TTS and Voice Cloning",
          "url": "https://huggingface.co/blog/open-tts-leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "Best Open Source Self-Hosted Text-to-Speech Models in 2026",
          "url": "https://pinggy.io/blog/best_open_source_self_hosted_text_to_speech_models/",
          "checked": "2026-10-25"
        },
        {
          "title": "FunAudioLLM/Fun-CosyVoice3-0.5B-2512 model card",
          "url": "https://huggingface.co/FunAudioLLM/Fun-CosyVoice3-0.5B-2512",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "granite-speech-5-0-2b",
      "name": "Granite Speech 5.0",
      "maker": "IBM",
      "category": "stt",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": null,
      "languages": 6,
      "runtimes": [
        "Transformers",
        "vLLM"
      ],
      "repo": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b",
      "leaderboards": [
        {
          "board": "Open ASR Leaderboard, English average WER (version 21-10-2026, 11 sets incl. 2 private)",
          "value": 5.68,
          "unit": "% WER",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Open ASR Leaderboard, version 21-10-2026 (English, 11 sets)",
          "url": "https://huggingface.co/spaces/hf-audio/open_asr_leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "ibm-granite/granite-speech-5.0-2b model card",
          "url": "https://huggingface.co/ibm-granite/granite-speech-5.0-2b",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "fish-audio-drama-3",
      "name": "Fish Drama 3",
      "maker": "Fish Audio",
      "category": "tts",
      "licence": {
        "weights": "Fish Audio Community License",
        "commercialClass": "C*",
        "condition": "Commercial use only below an annual-revenue threshold; never as a hosted service.",
        "checked": "2026-10-25"
      },
      "commercialUse": null,
      "tier": "server",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": false,
      "languages": null,
      "runtimes": [
        "SGLang-Omni",
        "vLLM-Omni"
      ],
      "repo": "https://huggingface.co/fishaudio/drama-3",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1248,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Artificial Analysis Speech Arena leaderboard (live table)",
          "url": "https://artificialanalysis.ai/text-to-speech/arena?tab=leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "fishaudio/drama-3 model card",
          "url": "https://huggingface.co/fishaudio/drama-3",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "ming-3-0-tts",
      "name": "Ming-3.0-tts 0.5B",
      "maker": "inclusionAI (Ant Group)",
      "category": "tts",
      "licence": {
        "weights": "MIT",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "8gb",
      "alsoTier": [],
      "memory": null,
      "voiceCloning": true,
      "cpuOK": null,
      "languages": 7,
      "runtimes": [
        "Ming-3.0-tts repository (PyTorch)"
      ],
      "repo": "https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B",
      "leaderboards": [
        {
          "board": "Artificial Analysis Speech Arena (main board)",
          "value": 1191,
          "unit": "Elo",
          "asOf": "2026-10-25"
        }
      ],
      "sources": [
        {
          "title": "Artificial Analysis Speech Arena leaderboard (live table)",
          "url": "https://artificialanalysis.ai/text-to-speech/arena?tab=leaderboard",
          "checked": "2026-10-25"
        },
        {
          "title": "inclusionAI/Ming-3.0-tts-0.5B model card",
          "url": "https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    },
    {
      "id": "phonon-31m-en",
      "name": "Phonon 31M",
      "maker": "Gradium",
      "category": "tts",
      "licence": {
        "weights": "Apache-2.0",
        "commercialClass": "C",
        "condition": null,
        "checked": "2026-10-25"
      },
      "commercialUse": true,
      "tier": "cpu",
      "alsoTier": [
        "8gb"
      ],
      "memory": null,
      "voiceCloning": null,
      "cpuOK": true,
      "languages": 1,
      "runtimes": [
        "Ollama 0.40.11+ (Apple Silicon)"
      ],
      "repo": "https://huggingface.co/gradium/phonon-31m-en",
      "leaderboards": [],
      "sources": [
        {
          "title": "Speech models in Ollama (Phonon)",
          "url": "https://ollama.com/blog/speech-models",
          "checked": "2026-10-25"
        },
        {
          "title": "gradium/phonon-31m-en model card",
          "url": "https://huggingface.co/gradium/phonon-31m-en",
          "checked": "2026-10-25"
        }
      ],
      "checked": "2026-10-25"
    }
  ]
}