Skip to content
Audio

AzurLane GPT-SoVITS V5dev: 902 character voices for voice generation

2 min read

A set of weights for the popular GPT-SoVITS V5dev speech synthesis system, trained on Japanese and Chinese voice lines from the game Azur Lane. The model covers 902 ship characters and lets you generate speech in Russian, Chinese, or Japanese using a short audio clip (3 to 10 seconds) as a reference. The package includes GPT checkpoints (up to the 15th epoch) and SoVITS LoRA (a lightweight add-on for the base model to adapt voices) with a rank of 32. The release includes detailed evaluation results based on voice similarity and pronunciation accuracy metrics.

Key features

  • Massive dataset: trained on 50,759 audio recordings totaling approximately 138.3 hours.
  • Flexible combinations: option to use 15th-epoch GPT weights alongside the official SoVITS or specialized SoVITS LoRA weights.
  • Included benchmarks: detailed comparison of various epoch combinations using ERes2NetV2, WavLM, and word error rate metrics.
  • Export format support: ability to convert models to AR GGUF for running via the fast local gsv.cpp server.

What you can do

  • Voiceovers for mods and videos: generate dialogs and lines for fan projects using Azur Lane character voices.
  • WebUI integration: run local speech synthesis in the official GPT-SoVITS WebUI by selecting the required `v5dev` config.
  • API usage: deploy a fast local backend based on gsv.cpp with Vulkan or CPU support.

Example

python tools/export_gguf/export.py --kind ar --ckpt "azur-lane-all-v5dev-20261121-e15.ckpt" --out "azurlane-v5dev-e15.gguf" --f16 --self-test
Read moreDetails
• Dataset: 50,759 audio fragments (49,727 Japanese and 1,032 Chinese, 114 TB lines).
• Architecture: GPT-SoVITS V5dev (AR + SoVITS architecture, LoRA with rank 32).
• Evaluation: tested on 16 voices, 24 Japanese, and 16 Chinese test phrases.
• Source code: based on the upstream GPT-SoVITS commit 6d51ffceafbf.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.7 GB
  • FP8 / INT8recommendedweights 0.1 GB0.6 GB
  • 4-bitweights 0.0 GB0.5 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 2060

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.

    # Moear/AzurLane-GPT-SoVITS-V5dev
    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 Moear/AzurLane-GPT-SoVITS-V5dev --local-dir ~/models/AzurLane-GPT-SoVITS-V5dev
    # 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 GPT-SoVITS (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 5All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

Moear/AzurLane-GPT-SoVITS-V5dev
Author
Moear
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments

Audio

pingo-ptbr

pingo-ptbr