Skip to content
Other

VoxCPM2 distilled indicF5

голос для индийских языков

1 min readupdated

Это дистиллированная TTS-модель VoxCPM2 на 2,29 млрд параметров, обученная на синтетическом телугу от sw-voice/indicF5-best. С 8 ноября 2026 года она обслуживает production voice B для языков te, kn, ta, ml, mr, gu, de и ru; hi, en и ja остаются на прежнем production-голосе.

Ключевые особенности

  • Поддерживает телугу, каннада, тамильский, малаялам, маратхи, гуджарати, немецкий и русский.
  • База — sw-voice/voxcpm2-indic-multi-stage2-english-reuse-lr1e-4-10ep.
  • Reference-free режим: cfg 2, 10 шагов, целые ответы за один проход.
  • Обучена на 5,6 ч синтетического телугу: 309 ответов и 202 письменных фрагмента.
  • Дистилляция заняла 320 шагов и около 21,5 минуты на одной H100.

Пример

Сценарий из карточки — синтез целых ответов без reference-аудио: cfg 2 и 10 шагов, один проход на ответ. Именно этот чекпойнт с 8 ноября 2026 года отдаёт production voice B для перечисленных языков.

ПодробнееСудья Gemini на целых ответах предпочитает модель учителю indicF5-best: +0,25 eval и +0,38 Q&A; против best-vox-2 — около +0,75.
Q&A CER — 0,96% (MAI-Transcribe-2); на трёх реальных текстах Navaratri — 0,8–1,6% против 2,3% у реальных записей.
Совпадение голоса — 95,0% против 98,3% у best-vox-2. Разброс высоты тона — около 3,7 полутона против 4,1–4,7 у реального голоса, примерно на 1,3 полутона ниже.
Discuss Telegram 0 19

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # sw-voice/voxcpm2-distilled-indicf5-te-synth5h
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download sw-voice/voxcpm2-distilled-indicf5-te-synth5h --local-dir ~/models/voxcpm2-distilled-indicf5-te-synth5h
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of voxcpm2-distilled-indicf5-te-synth5h (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

sw-voice/voxcpm2-distilled-indicf5-te-synth5h
Author
sw-voice
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

190

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments