Downloads · 30 days
57
36% of all-time downloads
DaRealDahoodie/dark-blues-v2-xl-vocals
dark-blues-v2-xl-vocals is a machine learning model from DaRealDahoodie. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft.
LoRA adapter for ACE-Step 1.5 XL SFT focused on expressive male vocal performances in a dark blues / noir jazz cinematic style.
Downloads · 30 days
57
36% of all-time downloads
All-time downloads
157
Public
Repo size
168 MB
Likes
0
Public
Click a slice to open those files.
.safetensors168 MB · 100%
From the Hugging Face model README
LoRA adapter for ACE-Step 1.5 XL SFT focused on expressive male vocal performances in a dark blues / noir jazz cinematic style.
Companion adapter: DaRealDahoodie/dark-blues-v2-xl-instrumental (instrumental version)
This adapter targets vocal stems with a moody, cinematic, jazz-inspired male vocal performance — late-night reflection, controlled tension, and the dark blues / noir jazz aesthetic.
It was trained using the "lora-balanced" recipe for good vocal quality and expressiveness while maintaining the core stylistic fingerprint.
Primary use: Generating dark, atmospheric tracks featuring expressive male vocals in the dark blues / noir jazz / cinematic vein.
Note: This is a very stylistically narrow adapter (the source material was intentionally focused). It excels at this specific dark blues / noir jazz cinematic vocal style rather than broad generalization.
acestep-v15-xl-sft model.adapter_model.safetensors).vocal stem, expressive male vocal performance, cinematic, jazz, dark blues, noir jazz, moody, controlled tension, late night reflectionSee the official ACE-Step 1.5 LoRA Training Tutorial for exact loading steps in the app.
vocal stem, expressive male vocal performanceadapter_model.safetensors — the LoRA weightsadapter_config.json — PEFT configurationMIT License
Copyright (c) 2026 PiMPStudios
See the LICENSE file for the full license text.
This is consistent with the base model ACE-Step/acestep-v15-xl-sft.
Created with: PiMPStudios/anvil-audio-v2 + ACE-Step 1.5 training stack (cloud job).
Best checkpoint selected: epoch_90_loss_0.5014 from the lora-balanced run.