Skip to content
Audio

Konkani Orpheus TTS

LoRA-адаптер для речи на роми-конкани

1 min read

Это LoRA-адаптер rank-64 от Reubencf для text-to-speech на базе unsloth/orpheus-3b-0.1-ft. Обучен на синтетической речи на роми-конкани; базовые веса отдельно и здесь не дублируются. Предназначен для одного синтетического голоса Konkani.

Ключевые особенности

  • Адаптер rank-64 поверх Orpheus-3B, библиотека peft.
  • Обучение: 18 300 клипов, 4 575 шагов оптимизатора, одна эпоха.
  • Один синтетический голос; транскрипты машинно скорректированы.
  • Для прямого инференса нужен служебный conditioning-токен fola: перед размеченной транскрипцией; приложения добавляют его автоматически.
  • Валидационные и тестовые примеры не входили в обучение.

Пример

Сценарий из карточки: загрузите базовую модель unsloth/orpheus-3b-0.1-ft вместе с адаптером и передайте токен fola: перед размеченной транскрипцией на роми-конкани. В приложениях токен подставляется автоматически.

ПодробнееОбучение: 18 300 клипов, 4 575 шагов, 1 эпоха, rank-64.
Валидация и тест не использовались в обучении; чекпоинт выбирали не по validation loss.
В обучении сохранены vocal-event теги; выразительность и произношение требуют проверки на слух.
Базовые веса не входят в репозиторий — нужна отдельная загрузка.
Discuss Telegram 0 88

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 6.5 GB · ready files7.7 GB
  • FP8 / INT8recommendedweights 3.4 GB4.3 GB
  • 4-bitweights 2.1 GB2.8 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 5060 Ti 8GB

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

This is a LoRA: counted together with its base model orpheus-3b-0.1-ft.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # Reubencf/konkani-orpheus-tts
    # ===== 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 unsloth/orpheus-3b-0.1-ft --local-dir ~/models/orpheus-3b-0.1-ft
    hf download Reubencf/konkani-orpheus-tts --local-dir ~/models/konkani-orpheus-tts
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to orpheus-3b-0.1-ft
  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

Requires the base model

unsloth/orpheus-3b-0.1-ft

This model is a LoRA / adapter. Download the base model first, then load this one on top.

  1. Llama-3.2-3B-Instructfamily root
  2. orpheus-3b-0.1-ft
  3. orpheus-3b-0.1-ft
  4. konkani-orpheus-tts

New posts about versions of Llama-3.2-3B-Instruct (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 6All on Hugging Face
Merges 1All on Hugging Face

About the model

Reubencf/konkani-orpheus-tts
Author
Reubencf
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

880

Ask AI about this model

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

Comments