Modelpublic

DisturbingTheField/ACE-Step-v1.5-acoustic-guitar-and-a-merge-LoRA

sourceHugging Facemitupdated 7mo agoView on Hugging Face

1likes63downloads

Model Card

Date: 2/23/2026

Rename these LoRA files to only adaptermodel.safetensors when using. The same adapterconfig.json can be used for both LoRA files.

LoRA file: adapter_model.safetensors

Sounds: Acoustic guitar, finger style playing Training: 6 self recorded wav files, 84 MB, 500 epochs, 45 minutes training LoRA Scale recommended: 0.2 - 0.7

LoRA file: vocalinstrumentmergeadaptermodel.safetensors

Sounds: 3 LoRA adapters merged. Instrumental songs by me, hoarse singing by me and acoustic guitar by me LoRA Scale recommended: 0.2 - 0.7

Both trained with acestep-v15-base and acestep-5Hz-lm-4B. Dataset, preprocessed tensors with both, and training with only acestep-v15-base. Both trained on a laptop, Nvidia RTX 3080 Ti with 16 GB VRAM. When making songs I used acestep-v15-turbo with acestep-5Hz-lm-1.7B.

NOTE: You can use the Python script MERGE-LORA.py.txt for merging two LoRAs. The LoRA vocalinstrumentmergeadaptermodel.safetensors is a merge of 3 LoRAs. First hoarse singing LoRA with instrumental LoRA. Then this merged LoRA with this new acoustic guitar LoRA. You can use the same adapter_config.json with all LoRAs.

General LoRA Information

Here is a new LoRA that contains acoustic guitar playing, fingerstyle, recorded by myself. There were 6 WAV files totaling approximately 84 MB. Trained for 500 epochs. Use this with a LoRA scale between about 0.2 and 0.7. You may get somewhat better acoustic guitar sounds, but possibly also a darker and calmer style.

I have also created a combination of this one with two earlier LoRAs—one with raspy singing and one with some instrumental tracks. That allows for even more adjustments to the songs you create: slightly raspier and less “perfect” vocals, some changes to instrument sounds and the guitar. The songs may also become a bit calmer, and perhaps less “country” than what is described in the caption.

You’ll just have to experiment and see what you get. You can also create other combinations yourself by using the script that merges LoRA files.

1 caption for testing:

Contry/folk music: A heartfelt country ballad built around fingerstyle acoustic guitar, warm bass, and steady, understated drums. The acoustic guitar carries the song with delicate picking patterns that create both rhythm and melody, while subtle steel-string slide guitar adds emotional swells and gentle accents. The bass provides a deep, supportive foundation, and the drums maintain a restrained, slow groove that allows the story to unfold naturally. A male vocalist delivers the lyrics with warmth and sincerity, his voice rich in the lower register and filled with reflective emotion. The overall atmosphere is intimate yet full-bodied, blending traditional country storytelling with a tender, nostalgic tone.

Lyrics for testing:

[Intro] Front porch light still shining on Even though the kids are gone Wind chimes singing soft and slow In a house we used to know

[Verse 1] There’s a drawing on the fridge Faded now along the edge Little handprints on the wall Marked the years so small, so small Matchbox cars and paper crowns Echoes in these empty rooms now

[Chorus] When did time slip through our hands? Just yesterday we made our plans Running circles through these halls Now it’s just these quiet walls It’s you and me where laughter used to be Standing in the space our babies used to fill

[Verse 2] Your coffee cup across from mine We sit and watch the evening sky Talking ‘bout the days gone by First steps and lullabies Every scraped knee, every tear Feels like they were just here

[Chorus] When did time slip through our hands? Just yesterday we made our plans Runing circles through these halls Now it’s just these quiet walls It’s you and me where laughter used to be Holding on to all we built

[Outro] Front porch light still shining on Guess love’s the reason we’re still strong The house is quiet, that is true But it still holds me and you And all the years we’ve traveled through In every room, our love grew

Model Description:

If you use "startgradioui.bat" then edit the file: set INITVICE=--init_service false So you get access to the LoRA loading part in the Web UI.

Model Sources: https://huggingface.co/ACE-Stepr/AceStep1.5

LoRA file: adapter_model.safetensors

LoRA file: vocalinstrumentmergeadaptermodel.safetensors

General LoRA Information

1 caption for testing:

Lyrics for testing:

Model Description:<br>