Dark Techno Sub-Bass Rumble (ACE-Step 1.5 XL)

# Dark Techno Sub-Bass Rumble LoRA (ACE-Step 1.5 XL)

This LoRA is explicitly trained to generate brutal, dark, and underground industrial techno tracks with a heavy focus on deep, synthesized sub-bass rumbles and driving, pounding kick drums.

Perfect for creating warehouse techno club vibes, industrial rhythm elements, and dark, hypnotic electronic music.

---

### 📊 Model & Training Details

- Base Model: ACE-Step v1.5 XL (SFT-XL)

- Dataset: 343 high-quality 30-second stereo slices (44100Hz)

- Target Style: Heavy sub-bass rumble, brutalist acoustics, industrial warehouse vibe.

- File to Upload: adapter_model.safetensors (Located inside the /adapter folder of your epoch checkpoint. Rename to your liking, e.g., Dark_Techno_Rumble_XL_v1.safetensors).

---

### 📈 Training Curve Analysis

Loss

1.0 ┤ \

0.8 ┤ \ /\_/\ /\_

0.6 ┤ \____/─v─v─\_______/───v── (Loss smoothed stabilized ~0.64)

0.4 ┤

└─┬─────┬─────┬─────┬─────┬─────┬── Step

0 250 500 750 1000 1500 2000

- Fast Convergence: The model converges rapidly within the first 300 steps.

- Flat Plateau: The smoothed loss stays perfectly locked between 0.60 and 0.65 across the entire run.

- Zero Clipping: This flat plateau guarantees stable sub-bass generation, preventing any digital distortion or "deep-frying" in the rumble frequencies.

---

### 🎙️ Trigger Words & Prompting Guide

To unlock the full power of the dark techno engine, always include the core style tags. You can dynamically adjust the speed (BPM) and scale (Key) at the end of the prompt.

#### ⚡ Core Style Tags (Copy all for maximum rumble):

dark techno style, heavy industrial rumble bass, sub-bass rumble, pounding driving kick, dark hypnotic atmosphere, industrial warehouse vibe, metallic percussion, gritty synths, underground techno, high quality

#### 🎵 Dynamic Tags (BPM & Key Examples):

- 135 bpm, key of G Minor

- 138 bpm, key of A Minor

- 140 bpm, key of F Sharp Minor

#### 📝 Full Prompt Example:

> dark techno style, heavy industrial rumble bass, sub-bass rumble, pounding driving kick, dark hypnotic atmosphere, industrial warehouse vibe, metallic percussion, gritty synths, underground techno, high quality, 138 bpm, key of A Minor

---

### 🧩 Required ComfyUI Nodes & Modules

To run this model successfully and hear the output, you need the following pipeline setup:

1. ACE-Step Model Loader (Loads the base acestep-v15-xl-sft checkpoint).

2. Lora Loader (Audio) (Loads this trained .safetensors LoRA file).

3. ACE-Step Prompt / Conditioning (The text input box for your prompt).

4. Audio Sampler / Diffusion Sampler (The core inference engine).

5. ACE-Step VAE / Vocoder Decoder (CRUCIAL! Decodes the internal latent tokens back into a crystal-clear, playable stereo .wav file).

6. Audio Save / Preview Audio (To listen to and export your generated tracks).

#### 🔀 Complete Inference Signal Flow:

[Base Model + LoRA] ➔ [Prompt Conditioning] ➔ [Audio Sampler] ➔ [VAE / Vocoder Decoder] ➔ [Preview / Save Audio]

---

### 🎛️ Optimal Inference Parameters (Sampler Settings)

Use these native audio settings in your sampler node to get clean, club-ready bass frequencies without digital clipping:

- LoRA Scale / Weight: 0.75 (Sweet spot for clean integration)

- Steps: 50 to 100 (More steps yield cleaner sub-bass rendering).

- CFG / Guidance Scale: 4.0 to 5.5 (Sweet spot is 4.5 to prevent distortion).

- Sampler Name (Native Options): er_sde (Best stability) or euler (Punchier transients).

- Scheduler (Native Options): linear_quadratic (Recommended for clean rhythm) or simple.

- Duration: 30.0 seconds (Matches the exact training length).

- Sample Rate: 44100 Hz (Studio standard CD quality).

- Lyrics Input: Always type [inst] or [Instrumental] to prevent the model from generating unwanted vocal artifacts.