Prompts Q

The best format for prompts - The most reliable prompt format is: [Genre + Era], [Mood/Emotion], [Key Instruments], [Vocal Style], [Production Quality]. Keep it under 120 characters for the style field and put detailed structure in the lyrics field with metatags.

Descriptors included in a prompt - The sweet spot is 4-7 descriptors. Fewer than 4 produces generic results. More than 7 tends to confuse the AI and create inconsistent output.

AI ignores vocal instructions - Vocal instructions often get deprioritized. Put vocal descriptors first or most prominent in your prompt. Try "Female vocalist, breathy soprano" at the start rather than buried at the end. Also use negative prompts: "no male voices."

The most useful metatags - The core five: [Intro], [Verse], [Chorus], [Bridge], and [Outro]. These provide the most reliable structural control. Additional useful tags include [Pre-Chorus] for builds and [Instrumental] for solo sections.

To generate instrumental music only - Include "instrumental only" or "no vocals" in your style prompt. You can reinforce this in the lyrics field by adding descriptive text like "(ambient synth pads)" instead of actual lyrics.

For songs not to start the same way - Break default to common patterns by specifying intros explicitly: [Intro: solo piano, 8 bars] or "Starts with isolated drums building to full arrangement." Vary your prompts between generations.

Regenerating before changing the prompt - Finding the exact vibe often takes 6+ attempts. If three generations miss completely, revise your prompt. If you're getting close but not perfect, try 5-6 variations before rewriting.

Character limit for prompts - On some platforms the style prompt has approximately a 200-character limit. Keep core style prompts concise and use the lyrics field for detailed structural instructions with metatags.