Reference-Track-to-Prompt Extractor for AI Music Generators

saas weekend hack • single request

Suno/Udio/Mureka users keep wanting the same workflow: drop in a song they like, get back a structured prompt they can paste into any model — describing genre, BPM, instrumentation hierarchy, mood, vocal character — without having to listen and transcribe by ear. Suno's own reference upload is a black box; an external, model-agnostic prompt extractor is missing.

builder note

Skip multi-iteration LLM-describes-music-then-LLM-recreates loops (the top reply correctly notes this loses essence). Use a small audio embedding model (CLAP, MERT) to extract tag distributions, then template-fill prompts in Suno/Udio/Mureka dialect. Per-platform prompt fidelity is the moat.

landscape (3 existing solutions)

AI music gen is fragmenting (Suno, Udio, ACE-Step, Mureka, Lyria) and creators want one neutral analyzer. Audio-tagging models are commodity now; the missing piece is the prompt-format adapter for each target model.

Suno reference track upload Closed-loop — works only inside Suno; you don't get back a portable prompt to use with Udio, Mureka, ACE-Step, etc.

Soundverse / Mureka style controls Manual sliders and text — no inverse 'analyze this MP3' shortcut.

Open-source MIR libraries (Essentia, librosa) Researcher tooling; nothing packaged as a click-to-prompt app for music creators.

sources (1)

reddit https://www.reddit.com/r/SomebodyMakeThis/comments/1suee8b/c... "Upload a song and get a rough breakdown you can turn into prompts. Faster way to capture the vibe." 2026-04-19

ai_musiccreator_toolsaudio_analysisprompt_engineering