Search Your Own Song and Watch Someone Else Win
Try this experiment with any independent artist: search their song title plus the word "lyrics." More often than not, the top result is not the artist. It's a lyrics aggregator - a site that never touched the session files, never heard the vocal stem, and in a lot of cases never had a human listen to the song at all. The words on that page are a guess. Sometimes a crowd-sourced guess, sometimes a scraped copy of someone else's guess. And because it ranks, that guess becomes the public record of what you said.
I decided that was not going to be the public record of my catalog. So dajai.io now carries its own lyrics archive - built from the masters, corrected by my own ears, and labeled honestly about which pages have been through that process and which haven't.
How a Page Gets Made: Masters In, Whisper Drafts, My Ears Decide
The pipeline has two passes, and the order is the whole point.
Pass one is machine transcription from the source audio. Not a streaming rip, not a YouTube capture - the actual master vocals from my own catalog. Whisper runs over that audio and produces a timestamped draft with a confidence reading on what it heard. That draft is a skeleton, nothing more. Speech models are decent at plain English and terrible at everything that makes a rap vocal a rap vocal: slang, ad-libs, deliberate misspellings, doubled takes, words that only exist on my records.
Pass two is me. Every line in the draft gets checked against the recording by the person who wrote it. Wrong words get fixed. Ad-libs get restored. And here is the rule that makes the whole thing worth doing: no invented bars, ever. If a line can't be confirmed against the audio, it gets flagged as uncertain - it does not get filled in with a plausible guess. A lyrics site guesses because it has to. I don't have to. I have the masters and I have the memory of writing it.
Only after that audit does a page publish. That's the difference between a transcript and a source-verified document.
Two Badges, Zero Pretending
Not every page in the archive has been through my hands yet, and the site says so out loud. There are two tiers, and every lyric page wears its tier as a badge:
- Emerald - VERIFIED / SOURCE-CORRECTED. These pages were transcribed from the master vocals and then hand-corrected by me, line by line, against the record. The verified tier contains no scraped lyrics. When you read an emerald page, you are reading the canonical text of that song.
- Amber - CATALOG / GENIUS-SOURCED. These pages were imported from my own artist page on Genius, so the words came from the community-lyrics world rather than from my audit. They're in the archive because having my catalog findable on my own domain matters, but each one carries a plain-language note that it hasn't been hand-verified yet. As I work through the backlog, amber pages get audited against the masters and promoted to emerald.
Most lyrics sites present every page with the same implied authority, whether a human ever checked it or not. The two-tier system is the opposite of that. Provenance labeling isn't a disclaimer I was forced into - it's the feature. You always know exactly how much to trust the text in front of you.
The archive is growing on both tiers, with more tracks sitting in the transcription queue - the count moves weekly, so the index is always the current picture. Flight to Vegas is already in the pipeline behind R2S; more on that below.
Why the Page Has to Live on My Domain
This is an ownership argument, the same one that applies to masters and stems. When a third-party site ranks for "your song + lyrics," you've donated three things: the traffic, the accuracy, and the authority to define your own text. Every model and search engine ingesting the web reads whatever version it can find - and if the findable version is a guess, the guess gets embedded as truth.
Every lyrics page on dajai.io emits structured data that ties the recording to my canonical artist entity, and the lyric text never enters that structured data as a placeholder - if a page doesn't have the real text yet, the schema simply carries no lyric text at all. Each page also wears its provenance badge in plain sight, so you always know whether the text has been through my audit. The goal is simple and unapologetic: search any song of mine plus the word "lyrics," and my domain should be the answer. Yours should be too, for your catalog. Syndicate outward to the aggregators if you want - but the source of truth should sit on ground you own.
Read Along: R2S Greatest Hits
R2S Greatest Hits is the compilation where the verified tier started. Put the album on, open the pages, and read along with the record - every emerald page below was corrected against the master vocals before it went live.
Start at the album page: R2S Greatest Hits.
Then read along, track by track:
- Far Side
- Hottie
- Wynn
- Flexico
- Mafi
- Get Rich
- On Go
- Did a Lot
- Going Up
- Some Thangs
- Put It Up
- Like This (feat. 1TakeJay)
For everything else, the index pages are the front door: dajai.io/lyrics lists every page with its badge and tier stats, and dajai.io/genius is the annotated knowledge layer built on top of the same archive.
Next Up: Flight to Vegas
Flight to Vegas is the next album through the exact same pipeline. The tracks are moving through the transcription queue now - Whisper first pass from the master vocals, then my line-by-line audit before anything wears the emerald badge. As pages clear, they land on the lyrics index with their tier showing, and the Flight to Vegas project page is the album's home base in the meantime. Same rules as R2S: no invented bars, uncertainty flagged, provenance visible. When the full read-along list is live, it gets the same track-by-track treatment you see above.
FAQ
What does the emerald VERIFIED badge actually guarantee?
That the page was transcribed from the master vocals with Whisper and then hand-corrected by me, line by line, against the recording before publishing. No scraped text exists in the verified tier, and no uncertain line was ever filled in with a guess - uncertainty gets flagged, not invented.
Why publish the amber CATALOG pages at all if they aren't verified?
Because an honest, clearly-labeled import from my own Genius artist page on my own domain still beats an anonymous scraper page as the destination for my catalog. The amber badge tells you exactly what the page is - imported, not yet hand-verified - and each one is in the queue to be audited against the masters and promoted to emerald.
Why not just maintain lyrics on Genius or another lyrics site?
Those platforms are distribution surfaces, not a source of truth. You don't control the accuracy, the ranking, or how the text gets scraped and re-used. Build the canonical version on your own domain first, keep the provenance visible, and let the third-party sites be downstream of you - never the other way around.
Can I follow along with an album while streaming it?
That's the intended use. Open the R2S Greatest Hits album page, press play on your streaming service of choice, and keep the lyric pages open in another tab. Every verified page is the text as it was actually recorded - read along with confidence.