Hum — AI music generator

Product — chat (or paste anything) into a song.

Describe a vibe, upload a deck/photo, drop a GitHub link or YouTube URL — Hum writes style + lyrics and generates the track in one conversation.

hum.demo.video.mov

Live app → · Built for MiniMax Week — Synthesis Track

---

Why this product

Music generation is usually a form and a prompt box. Hum is a conversation: editable title/style/lyrics in chat, multiple input types (text, image, doc, repo, video), then one click to a finished song with lyrics embedded in the MP3.

How it works

Chat / upload → title + style + lyrics → Generate → play + download

Features

Area	What you get
Create	Chat prompts · 12 genre templates · song or instrumental
Inputs	Text · image · .docx/.pptx · GitHub repo URL · YouTube URL
Edit	Click-to-edit title / style / lyrics · regenerate variations
Output	Play in-browser · WAV/MP3 · lyrics in MP3 tags · download all
UX	Setup banner · verse/chorus layout · timers · responsive · keyboard

Limits (honest): full-song generation only (no partial vocal re-record). Each Generate is a new take.

Use cases

Use	How
Team anthem / all-hands	Paste the quarterly deck
Content & social	Image or caption → short track
Open-source / dev community	Paste your GitHub repo
Learning	Notes → lo-fi study track
Personal	Travel photos, resumes, in-jokes → a song

Sample track

• demo/hum-of-creation.mp3 — showcase output from the app

Run your own

1. API key — GMI Cloud or MiniMax
2. Proxy — proxy/ on Render Blueprint (or python proxy/app.py locally)
3. Frontend — this repo’s index.html (GitHub Pages or any static host)
4. Configure — provider + key + proxy URL in the app

cd proxy && pip install -r requirements.txt && python app.py
# frontend: open index.html or serve the repo root

Repo layout

Path	Role
index.html	Single-file frontend
proxy/app.py	CORS proxy + lyrics embedding (Flask)
render.yaml	Render deploy config
demo/hum-of-creation.mp3	Sample generated track

Tech

Vanilla HTML/CSS/JS · Flask proxy · MiniMax-M3 (chat/lyrics) · MiniMax Music3.0 (audio) · GitHub Pages + Render · Lucide · Inter

Privacy: API keys stay in sessionStorage; generation goes through your proxy to the music provider you choose.

Family

• mimo-reader — browser TTS product
• hanna — Chrome TTS extension
• lux-tts — Colab voice-clone demo

License

MIT · datafying