MLTNT Militant Roots Reggae LoRAs for YuE2

MLTNT Militant Roots Reggae LoRAs for YuE2

Steers the YuE2-3B music model toward modern militant roots reggae with six LoRA files, from steppers to dancehall and hip-hop fusion. For musicians and tinkerers who want songs from lyrics.

Made by becausereasons on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Six LoRA files that steer the YuE2-3B music model toward modern militant roots reggae, with raspy patois male vocals, steppers grooves, sub bass, Hammond organ, nyabinghi drums and horn stabs. Each file patches both parts of YuE2 at once, the planner that writes the score, arrangement and vocal lines, and the decoder that makes the sound. One trigger word, mltnt, switches the style on for all six.

Who it suits: Musicians, producers and tinkerers who want reggae songs generated from lyrics and who are happy to tune a few settings per file.

Industry | Who uses it | How they use it | Example
Music production | Producers making reggae demos | Generate a rough roots or steppers track to build from | A steppers draft with a big anthemic chorus
Songwriting | Lyricists | Hear a lyric in dancehall, lovers rock or hip-hop styles | Run one lyric through Soundclash at planner strength 0.5
Sound design | Media creators | Create style-matched background music | A slow roots cue with horn stabs and dub sirens
Research | Audio ML researchers | Study how a LoRA patches a planner and a decoder | Compare one prompt at clip 1.0 and clip 0.5

You can skip it if you need commercial use, since the license is non-commercial, or if you want a genre other than reggae and its hip-hop and dancehall neighbours.

Where to use it: The card points to ComfyUI and to a browser demo on Hugging Face.

• ComfyUI, with the card's core workflow that uses built-in nodes only
• ComfyUI, with the card's FS_Audio Suite workflow
• The MLTNT demo Space in a browser, with all six LoRAs on free ZeroGPU
• Windows, macOS or Linux with Python, to download the files

Use case | Who | What they get back
Rapid-fire lyrics | Dancehall writers | Fast vocal delivery from Frontline, or from Soundclash at clip 0.5
Heavy roots tunes | Reggae producers | A slow, weighty roots track from Chanter at clip 0.5
Genre fusion | Hip-hop artists | Boom-bap over a one-drop groove from Fusion

How to set it up: You need the YuE2-3B base model, a place that can load LoRA files, and one of the six files. The easiest start is the demo Space, which needs no install.

• The YuE2-3B base model, which these LoRAs patch
• ComfyUI, if you want to run it yourself
• Python, for the Hugging Face download commands

1. Install the Hugging Face tools and download this repository's files.
2. Set up ComfyUI with one of the card's two workflows.
3. Load one safetensors file as the LoRA.
4. Put the trigger word mltnt in your prompt.
5. Set the planner and decoder strengths from the table below.

pip install -U huggingface_hub
hf download becausereasons/yue2-mltnt-militant-reggae

How to use it: Write your prompt with mltnt in it, pick a LoRA, and set two strengths. The clip strength controls the planner and the model strength controls the decoder. The card made its demos with seed 7, 32 steps, the dpm_2 sampler and the sgm_uniform scheduler.

Tool | What it does
mltnt_soundclash.safetensors | The versatile one for hip-hop, hard dancehall and lovers rock. Start at clip 0.5 and model 1.0 for fast styles, clip 1.0 for slow songs.
mltnt_chanter.safetensors | A settled signature voice. Clip 1.0 gives a fast 160 BPM feel, clip 0.5 gives a slow heavy roots tune, with model 1.0 either way.
mltnt_frontline.safetensors | The pick for rapid-fire lyrics. Start at clip 1.0, model 1.0 and cfg 1.0.
mltnt_fusion.safetensors | The widest palette, mixing reggae with hip-hop. Start at clip 1.0, model 1.5 and cfg 1.4.
mltnt_steppers.safetensors | The flagship, with tight steppers and a big chorus. Start at clip 1.0, model 1.0 and cfg 1.0.
mltnt_roots.safetensors | The purist, with the straightest roots sound. Start at clip 1.0, model 1.0 and cfg 1.0.

Example: Load Soundclash, give it a dense hip-hop lyric with mltnt in the prompt, and set clip 0.5 and model 1.0. The card's own run of this setup came back as a 3:02 song at 96 BPM with bass drops.

For newcomers and experts: A LoRA is a small add-on file that changes how a big model sounds without replacing it. The easiest first try is the demo Space in your browser, where all six LoRAs are ready to pick.

For experts, each file patches the autoregressive planner and the flow-matching decoder together, and the two strengths are set apart. The two newer files come from Ostris AI Toolkit, a different trainer with early YuE2 support. For fast or odd styles, drop their planner strength to 0.5. Lowering clip frees the planner to choose other arrangements, at the cost of the settled feel at full strength.

Write a lyric and a prompt with mltnt
Load one LoRA file
Planner writes score and vocal lines
Decoder renders the sound
A finished reggae song

Good to know: The card's demos land at quite different tempos, depending on the file and the clip strength.

Title: Tempo of the card's demos
Soundclash hip-hop: 96 BPM
Soundclash lovers rock: 73 BPM
Chanter at clip 1.0: 160 BPM
Fusion roots grid: 76 BPM

• Strengths: the two newer files need planner strength 0.5 for fast or unusual styles.
• Workflows: the ComfyUI bundle is a separate download that the card links.
• License: cc-by-nc-4.0, a non-commercial Creative Commons license.
• Cost to run: none named, the demo Space runs on free ZeroGPU, and no API key is mentioned.
• Data it reaches: it downloads the LoRA files from Hugging Face, and the demo Space runs in the browser.
• Not for: commercial use, since the license is non-commercial.

On Hugging Face, 13 November 2026

• Task: text to audio
• Downloads in the last 30 days: 0
• Likes: 27
• License: cc-by-nc-4.0
• Languages: en, jam
• Based on: m-a-p/YuE2-3B
• Tags: lora, yue2, music-generation, text-to-music, song-generation, reggae, roots-reggae, dancehall, steppers, comfyui, fs_audio, ai-toolkit

Links

• Model: https://huggingface.co/becausereasons/yue2-mltnt-militant-reggae
• Files: https://huggingface.co/becausereasons/yue2-mltnt-militant-reggae/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download becausereasons/yue2-mltnt-militant-reggae

Then follow the usage in its model card. Nothing in this kit runs by itself.