1. Workflow Overview

This workflow automates music production by integrating DeepSeek-R1 LLM for lyric generation and ACE-Step for audio synthesis. Ideal for creating custom background music or full tracks based on text prompts.

2. Core Models

• DeepSeek-R1-Distill-Qwen-7B

  • Function: Generates structured lyrics (e.g., verses, choruses) from text prompts.

• ACE-Step

  • Function: Converts lyrics and style prompts into music (melody, rhythm, chords).

3. Key Nodes

1. LmcqDeepLoader

   • Role: Loads DeepSeek-R1 model (requires ComfyUI-LMCQ plugin).

2. LmcqDeepGen

   • Role: Processes system prompts (e.g., "You are a musician") and user requests to generate lyrics.

3. MultiLineLyrics

   • Role: Formats lyrics with markers (e.g., **Chorus**).

4. ACEStepGen

   • Role: Synthesizes music from lyrics and style (e.g., "R&B").

   • Installation: Requires ComfyUI-ACE-Step plugin.

5. GenerationParameters

   • Role: Sets audio specs (BPM=202, sample rate=60, duration=15s).

4. Workflow Structure (Groups)

• Lyric Generation Group:

  • DeepSeek model loading, prompt input, and lyric generation.

• Music Generation Group:

  • ACE-Step model, audio parameter tuning, and preview.

5. Inputs & Outputs

• Inputs:

  • Music style (e.g., "{R&B}").

  • Lyric theme (e.g., "lyrics about rain").

  • Audio parameters (BPM, duration).

• Output:

  • Music file (.wav/.mp3) ready for playback or export.

6. Notes

1. Dependencies:

   • Download DeepSeek-R1 and ACE-Step models to plugin folders.

2. Hardware:

   • GPU ≥12GB VRAM recommended. Run lyrics and music steps separately if needed.

3. Tips:

   • Specific style prompts (e.g., "90s retro R&B") improve output quality.

   • Reduce BPM/duration in GenerationParameters if audio stutters.