One call turns an idea into finished lyrics, a producer-grade style prompt, and a title. Built for anyone making music — whether a human writes the melody or an AI does.
Private beta · free sandbox to build against · first call in under 10 minutes
# one idea in → curl https://api.hookgenius.app/v1/generations \ -H "Authorization: Bearer hg_live_…" \ -H "Content-Type: application/json" \ -d '{ "idea": "a defiant comeback anthem", "genre": "arena rock", "vocal": { "gender": "female" } }'
{
"status": "completed",
"output": {
"title": "Teeth First",
"style_prompt": "arena rock, thunderous
drums, powerful female belt,
anthemic defiant energy…",
"lyrics": "[Verse 1]\nThey said
pray for me at the dinner table…"
},
"usage": { "credits": 1 }
}
Every request returns three things, each tuned to be used immediately: the words, the sound, and the name.
Complete, sung-ready lyrics with section tags, vocal direction, and structure — verse, chorus, bridge, the whole song.
A precise, comma-separated sonic brief — genre, instrumentation, vocal timbre, energy, production — that makes a generation engine actually sound the way you meant.
A real title that fits the song — not "Untitled Track 4."
Lyrics
[Verse 1] Left the porch light burning like a debt I won't repay Mom's wind chimes still talking to a yard that's gone to gray I know every pothole by its first and middle name [Verse 2] The water tower got my class year and a crown Seventeen coats of paint can't hold a dead king down
Anyone can send a prompt to a language model and get back something that rhymes. That's not what this is. Every generation runs through a tuned, multi-stage pipeline — it drafts several takes, critiques them against what actually makes a song land, and produces a final the way a real session would. The result is the difference between "an AI wrote this" and "this is good."
A single take is a coin flip. The engine generates multiple drafts and selects the strongest — reliability of quality per call, not luck.
A critic stage measures each draft against real songcraft — hook strength, coherence, the AI-tell — and refines before it ever reaches you.
Years of tested knowledge about how to phrase a sonic brief so a music engine renders it right. That domain expertise is the moat — and you get it in one field.
The best songwriting partner shouldn't care how you make the song. It just makes the song better.
Async by design — a great song takes a moment to write. Fire a request, get a job, receive the result by webhook or poll.
# 1 — create a generation (returns immediately) POST /v1/generations → 202 { "id": "gen_8f2…", "status": "queued" } # 2 — poll, or set a webhook, or long-poll with ?wait=60 GET /v1/generations/gen_8f2… → { "status": "completed", "output": { … } } # 3 — use it. title + lyrics + style_prompt, ready to paste anywhere.
You pay per finished song. No subscription to start, no minimums, no charge when a generation fails.
Prepaid · 1 credit = 1 finished song
Buy credits up front. Volume pricing kicks in automatically as you scale — down to $0.69.
Sign in with your email, generate a key, and prepay for credits — all self-serve. Build against the free sandbox before you spend a cent.
Open the console → Read the docsPut it in your product, your studio, your workflow. One call away.