Draft agreement and privacy notice

What this document is

This sets out what you are permitting when you contribute one song to the InnaWords research pilot. It is written in plain language on purpose. It is a draft prepared by the researcher, it is not legal advice, and it has not been reviewed by a lawyer.

Version 0.1.0-draft. If this text changes, the change applies to signatures made afterwards. It does not reach back and alter a permission you already gave.

Who you are agreeing with

Stefen Ewers, acting personally, as an independent researcher. Not a company, not a university department, and not on behalf of any institution. You can reach him at [not yet confirmed].

This pilot is not sponsored by, approved by, or affiliated with Georgia Tech, any record label, any publisher, or any streaming service.

What this covers

One specific recording, the one you name and upload in this submission. It does not cover your catalogue, your other versions, or anything you send later. Each track is a separate decision and a separate signature.

You keep ownership

You keep every right you had before you signed. Nothing here transfers ownership of the recording, of the words, or of the music. You may keep licensing, releasing, selling and performing the track exactly as you would otherwise.

What you grant is a limited, non exclusive, non transferable permission to use this one track for the research uses listed below. It is not a sale.

What you are permitting

By signing, you permit the researcher to do the following with this track, and nothing beyond it:

• Store the recording and the lyrics you provide in private storage that only the researcher can open.
• Run existing speech to text software on the recording to measure how accurately it transcribes your words.
• Align your lyrics to the recording and correct transcription errors by hand.
• Record notes about Patois spelling, pronunciation and meaning, kept separate from your original lyrics.
• Report combined results across all contributed songs in a form that does not identify you and contains no audio and no lyric excerpts.

Optional permissions

These are separate choices. Each one starts switched off. Declining any of them does not affect the rest of your participation, and you can take part fully without granting either.

• Adapting or fine tuning a speech to text model using this track. This is model training, not testing. Testing measures what existing software already does. Training changes the software using your recording. You can decline this and still take part in the evaluation.
• Publicly thanking you by artist name. Your artist name only. This does not publish your song name, your lyrics, any clip, or any result tied to you.

What is excluded

These uses are not permitted under any version of this agreement:

• No generating music, vocals or instrumentals.
• No voice cloning and no imitation of any artist.
• No selling, licensing or distributing your recording.
• No public dataset containing your audio or your lyrics.
• No public release of model weights trained on your track.
• No commercial use.
• No sublicensing to another company for their own purposes.
• No use of your track to promote anything.

This is unpaid

There is no payment, no royalty, no fee and no budget to pay one. Taking part is voluntary.

There is also no promise of exposure, of publication, of placement anywhere, or that any result will be useful to you. If any of that would be your reason for contributing, this is the wrong project.

What you can expect to receive

If your track is accepted and completes review, the intended outputs are a reviewed lyric transcript and a short summary of the errors the transcription software made on your song. Timed caption files are experimental and are provided only where feasible, after a person has checked them.

No date is promised. This is manual work done by one person, and a submission being received is not a submission being accepted.

Where your music is kept

Your audio and lyrics are stored in a private Supabase project, hosted on infrastructure in [not yet confirmed]. Supabase and its underlying hosting provider are processors: they hold the data so the researcher can, and they do not get to use it.

Nothing you contribute is sent to Anthropic, ElevenLabs, any analytics service, or any other AI service. The assistant elsewhere on this website has no access to any of it.

Access is limited to the researcher. The storage is private, not a public link.

How long it is kept

Contributed audio and lyrics are kept for up to a period that is [not yet confirmed] from the date you sign, and then deleted, unless you withdraw sooner.

A minimal record that a permission once existed is kept beyond that: the fact of a signature, its version, its date, and the choices you made. That exists so both of us can show what was agreed, and it holds no audio and no lyrics.

Withdrawing

You can withdraw at any time, by signing in and withdrawing, or by writing to [not yet confirmed].

Further research use and any export stops immediately when you withdraw. Your stored audio, your lyrics and anything derived from them are deleted on the schedule set out in the privacy notice. Routine backups expire on their own cycle, which is also described there.

Two honest limits. Findings already published cannot be unpublished, although they contain no audio, no lyric excerpts and nothing identifying you. And because no model training happens in this pilot at all, there is no trained model to remove your track from; if that ever changes, it will require a new agreement and a fresh signature rather than a quiet reinterpretation of this one.

Your signature

Typing your full legal name and confirming your intent is how you sign. The time is recorded by the server when you submit, in UTC.

Verifying your email address shows that you control that inbox. It does not prove that you own the music. Your declarations about rights are what this agreement relies on, which is why they are asked so specifically.

This document does not claim that an electronic signature is enforceable in your jurisdiction. That question has not been assessed.

What this covers

How the InnaWords pilot handles what you give it. Version 0.1.0-draft. A draft, prepared by the researcher, not reviewed by a lawyer.

What is collected

Only what the research needs:

• Your email address, verified so you can sign in and get your record back.
• Your legal name and artist name, because a permission has to come from a named person.
• Your country, and your role on the track.
• A yes or no answer about being 18 or over. Not your birth date, and no identity document.
• Your answers about who holds rights in the track.
• The track details, the lyrics you type, and any notes you add.
• The audio file you choose.
• The record of your signature: the version you signed, your typed name, the server time, and the choices you made. The IP address and browser you signed from are kept as part of that evidence trail.

What is not collected

No birth date, no government identification, no payment details, no location beyond the country you type, and no tracking of you around the web. This feature sets no analytics and no session replay. Your lyrics, your audio and the contents of your agreement never appear in logs.

Where it goes

A private Supabase project in [not yet confirmed]. Supabase, and the cloud provider underneath it, act as processors. The researcher is the only person with access.

Nothing is passed to any AI service, any analytics service, or any advertising service. Nothing is sold, and nothing is shared with a label, publisher or platform.

How long it is kept, and how it is deleted

Audio and lyrics: up to a period that is [not yet confirmed] from signing, then deleted.

On withdrawal: research use and exports stop immediately. Stored files and anything derived from them are deleted within 30 days of your request.

Backups: routine backups are kept by the storage provider on their own rotation and expire on that cycle rather than being edited. Deleted material can persist in a backup until that rotation passes, which is why the deletion timeline is described as a window rather than as instant.

Kept afterwards: the minimal record that a permission existed, described in the agreement. No audio, no lyrics.

If you filled in the rights contact form instead

That form collects your email, the situation you described, and the track and artist name if you gave them. No audio. No signature. It exists so a conversation can happen, and it is deleted when that conversation is finished or on the same retention schedule, whichever comes first.

Asking about your data

Write to [not yet confirmed] to ask what is held, to get a copy, to correct something, or to have it deleted.