Skip to content
Audio

orpheus-3b-swc-swa06

голос для начитки Писания на суахили

2 min read

Это файнтюн Sunbird/orpheus-3b-tts-multilingual — TTS-модели на базе Llama-3.2-3B, которая генерирует аудиокоды SNAC 24 kHz. Модель на 3,3B дообучена на 96 часах студийной библейской начитки на суахили и предназначена для чтения Ветхого Завета на конголезском суахили (swc, SWC02).

Ключевые особенности

  • База — Sunbird/orpheus-3b-tts-multilingual: причинная LLM Llama-3.2-3B, выдающая SNAC-аудиокоды.
  • LoRA r=64 слита в веса; обучение — 3 эпохи. В папке adapter/ лежит тот же чекпойнт как PEFT-адаптер на 390 MB.
  • Чекпойнт swa06 обучали на SNAC-кадрах без дедупликации — по карточке, это дало самую разборчивую модель из семи прогонов проекта.
  • Основной голос — openbible_swa_0001; в корпусе есть ещё шесть дикторов.
  • Важное ограничение: тренировочное аудио — прибрежный/стандартный суахили (swh), а не swc; модель читает написание swc, но с прибрежным акцентом.

Пример

Быстрый старт из карточки:

pip install "transformers>=4.45" torch snac soundfile numpy
python infer.py --model danielgloo/orpheus-3b-swc-swa06 --text "Hapo mwanzo Mungu aliumba mbingu na inchi." --out gen_1_1.wav
ПодробнееCER на 178 отложенных стихах × 3 сида: 0.0665 у модели, 0.047 у человека, 0.076 у базовой модели. Сходство голоса с диктором (WavLM x-vector cosine) — 0.96.
В продакшн-рендере Бытия (1 511 стихов) доля провалов с первого раза — 1.8%; все стихи прошли за ≤3 попытки. Пропускная способность — ~725 токенов/с при batch 16 на одной GPU.
Держите repetition_penalty ≥ 1.1: при 1.0 модель зацикливается. Стихи длиннее ~100 текстовых токенов могут укорачиваться.
Лицензия весов — CC BY-SA 4.0 (share-alike); Orpheus-3B — Apache-2.0, SNAC — MIT.
Discuss Telegram 0 23

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 6.1 GB · ready files7.3 GB
  • FP8 / INT8recommendedweights 3.1 GB3.9 GB
  • 4-bitweights 1.7 GB2.4 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.

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.

    # danielgloo/swahili-drc
    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 danielgloo/swahili-drc --local-dir ~/models/swahili-drc
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to swahili-drc
  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

  1. Llama-3.2-3B-Instructfamily root
  2. orpheus-3b-0.1-pretrained
  3. orpheus-3b-0.1-pretrained
  4. orpheus-3b-tts-multilingual
  5. swahili-drc

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

danielgloo/swahili-drc
Author
danielgloo
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

230

Ask AI about this model

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

Comments

Audio

Swanada

Swanada