About Music I Want

Last updated: April 2026 · Written by Dan Cohen, founder

What this is

Music I Want is a music analysis platform. Every song in the library — 19,215 songs across 1,466 artists — gets analyzed for BPM, tempo, dynamics, mood, texture, and vocal style. Look up any song and you get the answer to what people actually search for: how fast is it, how loud does it get, what mood does it sit in, what does it feel like.

We also analyze each song on a set of dimensions other music platforms don't bother with — sudden changes, predictability, how the vocal sits against the production, whether the mix has harsh timbres. That's how you know a "chill" song from streaming actually stays chill, or why a song you thought was calm left you rattled.

Why it exists

Streaming services are optimized for the middle. They assume any song is fine to put next to any other. For a huge number of listeners, that isn't true. A song that erupts at the second chorus is not a calm song even if it's tagged "chill." A song with a whispered vocal that suddenly screams is not background music even if it's slow. The standard rating systems — genre, BPM alone, "energy," a single mood tag — don't capture what listeners actually experience.

I built this because I couldn't find it. Not from Spotify. Not from Apple Music. Not from any of the wellness playlists that promise "calm." Calm, it turns out, isn't a property a playlist curator can assign by ear. It's a property of how a specific song behaves — how far it swings, whether it lurches, what the voice does over the top. So I started building a system that records those properties, song by song.

How each song is analyzed

Every song in the library is analyzed along the dimensions documented on the Methodology page. Analysis is model-assisted — derived from each song's documented musical properties (published tempo, instrumentation, arrangement, vocal delivery, production) and reviewed by hand. We do not analyze audio signal. Nothing in the library is a wholesale AI dump — when I tried seeding the database with AI-generated commentary in early 2026, the output was templated and I pulled it. The result: a smaller, slower-growing database where every entry reflects real listening, not pattern-matched filler.

The methodology is documented publicly so you can decide whether you trust it. If you think an analysis is wrong, the Analyze a Song tool and the Request Artist form feed directly into review.

Who it serves

If you want a quick answer ("what's the BPM of this song?"), you get it in the page title. If you want depth (how does the song build, where does it peak, what does the vocal do, is it going to startle you), it's all there. The site works for:

Runners and fitness people matching BPM to stride or workout pace.
DJs and producers finding songs that mix into what they're working on.
Playlist curators building coherent moods without mid-list surprises.
Listeners whose ears notice everything — anyone for whom a sudden change costs something real — who need to know what a song is going to do before pressing play.
Parents picking music for kids where a sudden cymbal crash ruins bedtime.
Anyone who wants more information about a song than "energy: 0.68."

Who built it

My name is Dan Cohen. I'm the founder of The Hive, a small studio that ships internet tools with long time horizons and zero investor pressure. Music I Want is one of those tools. You can reach me on LinkedIn. If you think an analysis is wrong, if an artist you love is missing, if a feature would make the site more useful to you — write.

I am not a neuroscientist. I am not an audiologist. I am a person who builds careful tools because I want the internet to still be useful to people who don't fit the algorithmic middle. The analytical framework is informed by published work on audio engineering and sensory processing, but the analyses themselves are not clinical. They are listener-side guidance.

How this site makes money

Honestly and transparently, because anything else corrodes the thing itself. The site runs Google AdSense (pending approval at the time of this writing — see our Disclaimer). Guide pages that review hardware include a small number of affiliate links to eBay and, where we qualify, Amazon. Every affiliate link is marked with rel="nofollow" and a visible "affiliate" tag above the block. There are no subscription paywalls. There is no scraping or resale of user data. There are no email blasts without explicit opt-in. If revenue ever permits it, the ads will go.

What this site is not

It is not a streaming service. We link out to YouTube, Spotify, and Apple Music so you can actually listen — we don't host the audio. It is not a replacement for a clinician if sensory processing or misophonia seriously affects your daily life. It is not an authoritative final answer on how any song will make you feel — individual response varies, and the analyses are guidance, not verdict. If our Low-Intensity rating doesn't match your experience of a song, trust yourself over us.

How to use it

Analyze a Song — paste a song name. Get BPM, tempo, mood, and the full analysis.
Finder — dial in what you want (BPM range, mood, intensity, use case) and get a short list.
Library — browse the whole database with full filters.
Voices — read (and write) long-form essays about bands that matter to people.
Guides — deep dives on BPM and running, music and focus, classical intensity, and more. All guides.

Credits and corrections

Writing that includes factual claims about artists, songs, or sensory science is cited where citations exist. If you see something wrong, send a note via the contact page. We keep a changelog in the guide pages — each guide records the date it was last updated and what changed. The site itself is versioned in a git repo we maintain; we can point you at specific commits if you want to verify anything.

How content is produced

Different categories of content go through different production paths. We disclose them all because the integrity of the framework depends on knowing which is which:

Acoustic ratings — the five dimensions on every song are model-assisted — derived from each song's known musical properties (published tempo, instrumentation, arrangement, vocal delivery, production style) plus human review. We do not measure audio waveforms. The combination is conservative: where assessments disagree, the more cautious rating wins. The full process is documented on the methodology page.
Editorial guides — the long-form guide pages on conditions, use cases, and frameworks (sleep music, anxiety music, kids music, sensory overload at work, etc.) are written by Dan Cohen as the founding editor. Some supplementary artist guides are drafted by AI under editorial supervision and pass through a structured QA pipeline (a second AI critique pass) before human approval. AI-drafted long-form pages indicate this in the page footer for transparency.
Imagery — the abstract illustrations attached to song pages are generated by AI image models (Flux Schnell on Replicate). Prompts are computed from the song's own data: title, sensory profile, mood, era. Prompts explicitly forbid faces, text, logos, and band imagery. The illustrations are abstract editorial art, not reproductions of any existing artwork.
User-submitted content — fan stories, song descriptions, Voice essays, and image votes come from listeners. Every text submission passes through a moderation queue with human approval before publication. We do not solicit or accept copyrighted song lyrics; the form explicitly omits any lyric input.

How content is moderated

Every piece of user-submitted content goes through a review queue. The reviewer (human, currently the founding editor) checks for:

Spam, automation, or attempts to game the voting system
Reproduction of copyrighted material (lyrics, third-party essays, AI-generated text passed off as personal experience)
Hateful, defamatory, or harassing content
Personal information about identifiable third parties
Content unsuitable for a general audience

Approved content is published and credited to the submitter's chosen first name or alias. Rejected content does not appear on the site; we do not return the rejection rationale unless the submitter follows up. We retain rejected submissions for 30 days for audit purposes, then delete.

Data sources

The Service relies on a small number of external data sources, all of which are credited where their data appears:

MusicBrainz (operated by the MetaBrainz Foundation) — for canonical song, album, and artist identifiers and basic metadata. We respect the MusicBrainz license terms.
Cover Art Archive — for album cover art images displayed on song pages. Images remain the property of their respective copyright holders and are loaded directly from the Cover Art Archive's CDN; we do not rehost them.
Spotify, YouTube, Apple Music — for embed players that let visitors listen to songs via those services under those services' own terms.
Our own analytical work — the five-dimension sensory framework, the use-case verdict matrix, the guide library, the editorial commentary, the methodology, and the relationships between songs (alternatives, similar by DNA, mood overlaps) are original.

Editorial principles

A few commitments that guide how we work:

Honest about uncertainty. Sensory analysis is not exact science. When we are confident in a rating, we say so plainly. When we are not, we say so plainly. AI-derived ratings are flagged as such.
Helpful before clever. The site exists to help listeners answer real questions about songs. Editorial flourishes that obscure the answer are cut.
No filler. A blank fact is better than padded prose. We would rather a shorter, denser page than a longer, watered-down one.
Specific over generic. “The song builds at 1:42 with a layered guitar entry” beats “the song builds dramatically.” We try to be specific even when it costs us pretty sentences.
Cite when claiming. Factual claims about artists, songs, or scientific findings carry citations or disclaimers when no citation is available.
Listener autonomy. Our ratings are guidance, not verdict. We trust your nervous system over our number; if our rating disagrees with your experience, your experience wins.

Accessibility

We try to make the Service usable by people with a wide range of abilities. Specific commitments:

Text-content pages render as plain HTML with semantic headings, lists, and paragraphs — not as JavaScript-only DOMs — so they work with screen readers and reader-mode tools.
All images carry alt text describing what they depict (album art, abstract illustrations).
Color contrast meets WCAG AA standards on all primary text. The dark theme is designed for low-stimulation reading and to reduce screen brightness in low-light environments.
Form fields have associated labels.
Embedded YouTube and Spotify players are loaded only when explicitly clicked, so users on metered or low-bandwidth connections are not auto-charged for media they did not request.
The cookie consent banner offers a clear binary choice rather than a dark-pattern wall.

If you encounter an accessibility issue, please contact us via the contact page with the words “Accessibility” in the subject line. We treat accessibility issues as urgent.

How we handle being wrong

We will be wrong. Sometimes the rating is wrong. Sometimes a fact is wrong. Sometimes a guide overgeneralizes. Our commitments:

Make corrections promptly when notified.
Note material corrections at the top of the affected page so readers know the page changed.
Keep guides dated, so readers can judge currency themselves.
Maintain a public changelog in our git history; we can point to specific commits if you want to verify anything.
Where a correction shifts a song's rating, we note both the old and new rating in the page's history.

What we don't do

Host audio — we don't stream or download music. We link to YouTube, Spotify, and Apple Music so you can listen on the original platforms.
Reproduce song lyrics — we don't display copyrighted song lyrics anywhere. The form fields that previously solicited lyrics from users have been removed; existing lyric data has been wiped.
Sell user data — full disclosure on data handling is in our Privacy Policy.
Run dark patterns — no cookie banners that hide “reject” behind a maze, no countdown timers on subscription forms, no fake-scarcity language.
Auto-play — no audio or video plays without your explicit click.
Push notifications — we do not request browser push permission.
Account requirements — you can use almost every feature without giving us anything. Email is required only for newsletter and saved-playlists features, and is opt-in.

Financial transparency

Music I Want is monetized in two ways:

Display advertising via Google AdSense. AdSense is the most common ad network on the open web; we use it to keep the lights on. Ad revenue scales with traffic. We do not influence which ads appear; Google's automated systems select them.
Affiliate links on guide pages, primarily through the eBay Partner Network. When a reader clicks an affiliate link to a recommended product (headphones, sensory gear, etc.) and purchases on the destination site, we earn a small commission at no extra cost to the reader. Affiliate links are visually marked with an “affiliate” tag and use rel="nofollow" so they do not influence search rankings.

We do not currently take payments. There is no subscription, no paywall, and no removed-after-X-days content. If we add any of those in the future, the change will be announced clearly in advance and existing content will not be retroactively paywalled.

We do not take venture capital. We do not plan to. The site is funded by a small individual proprietor, with the goal of being self-sustaining via ad and affiliate revenue without compromising editorial integrity.

The roadmap

The roadmap is shaped by what listeners ask for. Public ideas queue and recently-shipped notes live on the roadmap page. If something would make the site more useful to you — a new use case, a missing artist, an analysis dimension we haven't covered — suggest it there.

Credits and corrections

Contact: LinkedIn · Contact page · Methodology · Privacy · Terms · Disclaimer