ryanontheinside/ambient-acestep1.5-v1

Ambient

Ambient-style generation. Covers: ambient drone, evolving pads, ethereal textures, slow swelling, meditative; cinematic ambient score, sweeping pads, distant horns, emotional swells; and dark ambient, low drones, distant metallic textures, ominous atmosphere.

Quick start

• Base model: AceStep v1.5 Turbo (2B)
• Trigger word: roti-4mbnt (must appear in the prompt to activate)
• Recommended strength: 0.7
• Tested strength range: 0.3 – 1.0
• Recommended steps: 8
• Recommended guidance: 7.0
• Recommended shift: 3.0
• Sample rate: 48000 Hz

Example prompts

• ambient drone, evolving pads, ethereal textures, slow swelling, meditative (60 BPM, D minor)
• cinematic ambient score, sweeping pads, distant horns, emotional swells (70 BPM, F major)
• dark ambient, low drones, distant metallic textures, ominous atmosphere (50 BPM, C minor)

Training

• Rank 64, alpha 128, dropout 0.1
• Target modules: q_proj, k_proj, v_proj, o_proj
• 500 epochs / 7500 steps, batch 1, optimizer adamw_fused, lr 0.0003, warmup 100
• Dataset: 15 synthetic tracks, 3760s total, instrumental fraction 0.0

Files

• ambient-v1.safetensors — PEFT LoRA weights (88,130,248 bytes, sha256 bcd3e4869c38a5e66a80aa85ccc22e9edfdfbb4e25406b30fcdd63e4c40ca266)
• ambient-v1.metadata.json — machine-readable sidecar (schema v1)
• adapter_config.json — PEFT adapter config

Full metadata: see ambient-v1.metadata.json.