Skip to content
Audio

90s Trance LoRA для ACE-Step 1.5 XL SFT

длинные транс-треки из промпта

1 min readupdated

LoRA от LouisAmas для ACE-Step 1.5 XL SFT (4B DiT) сдвигает генерацию в 90s trance: uplifting, acid, goa, progressive и early rave на 130–142 BPM. Модель делает длинные DJ-интро, breakdowns и builds, а также 7-минутные инструментальные треки. Trigger word — ys_trance90s, его ставят первым в caption.

Ключевые особенности

  • Стили: uplifting, acid, goa, progressive и early rave; темп 130–142 BPM.
  • Длинные DJ-friendly интро, реальные breakdowns и builds; треки до 7 минут.
  • Только инструментал: lyrics [Instrumental], вокал не обучали.
  • Запуск на acestep-v15-xl-sft, не на turbo: turbo перебивает LoRA. Планировщик — acestep-5Hz-lm-4B.
  • Рекомендуемые настройки: LoRA strength 1.4, 50 steps, Euler, CFG 7, duration 420 s; BPM и тональность задавать явно.

Пример

ys_trance90s, 90s trance, 136 BPM, A minor,
uplifting, driving, bright, instrumental
ПодробнееОбучение: 40 инструментальных треков 90s trance (4,3 ч), только lossless FLAC, 24 бит / 48 кГц; LoRA rank 128, alpha 256.
Железо: 1 × GPU 48–96 ГБ, около 1,5 ч.
Ограничения: только инструментал, вокал не обучали; длинные треки могут зацикливаться; тренировочные треки не распространяются, перед коммерческим релизом проверяйте выход из-за copyright.
Discuss Telegram 0 20

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 19 GB · ready files21 GB
  • BF16weights 9.6 GB11 GB
  • FP8 / INT8recommendedweights 5.0 GB6.0 GB
  • 4-bitweights 2.9 GB3.7 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 · F32 — from RTX 4090

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 acestep-v15-xl-sft.

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.

    # LouisAmas/acestep-1.5-xl-sft-lora-90s-trance
    # ===== 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 ACE-Step/acestep-v15-xl-sft --local-dir ~/models/acestep-v15-xl-sft
    hf download LouisAmas/acestep-1.5-xl-sft-lora-90s-trance --local-dir ~/models/acestep-1.5-xl-sft-lora-90s-trance
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to acestep-v15-xl-sft
  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

ACE-Step/acestep-v15-xl-sft

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

New posts about versions of acestep-v15-xl-sft (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 1All on Hugging Face
Fine-tunes 7All on Hugging Face
LoRAs and adapters 3All on Hugging Face
Merges 1All on Hugging Face

About the model

LouisAmas/acestep-1.5-xl-sft-lora-90s-trance
Author
LouisAmas
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

200

Ask AI about this model

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

Comments

Audio

Dragonfly

Dragonfly