# Contributor Consent and Rights Declaration

**Project:** OpenSanta Project
**Version:** 0.4 draft (project named OpenSanta Project; CC BY 4.0 dataset license; required signed form + spoken consent clip). Not yet reviewed by a lawyer.
**Contact for questions or withdrawal:** [PROJECT CONTACT]

## What this project is

We're building a free, openly licensed "classic American Santa" text-to-speech (TTS) voice.
Volunteers record lines from a shared script. We publish the recordings as an open dataset
and use them to train a voice model that **anyone can use for any lawful purpose, including commercially.**

Please read this whole page before you record. If any part of it doesn't sit right with you,
don't contribute. That's completely fine.

---

## 1. Who can contribute

- **You must be 18 or older.** Age is **self-declared**: you confirm it with the "I am 18 or older" checkbox
  in the signed form and by saying it in your spoken consent clip. We don't ask for ID or any other age check.
  There's no parent or guardian path: we don't accept recordings of anyone under 18, even with permission.
  If you're submitting someone else's voice under the exception below, that speaker must also be 18+ and
  self-declare in their own consent clip.
- **Accent (v1):** we're collecting General American accents for one consistent classic American Santa voice.
  This isn't a judgment of any accent. We may decline submissions, or leave them out of v1 training, after
  listening to your core set, and we'll tell you if we do.
- **You must be the speaker in every recording you submit.** The only exception: you hold all
  rights to the audio **and** you have the speaker's own written consent to every item in
  section 3. Send that consent with your submission. We may still decline it.

## 2. What audio is NOT allowed

Only submit audio you recorded yourself for this project, or audio you fully own. Do **not** submit:

- **Scraped or ripped audio** from YouTube, TV, films, podcasts, audiobooks, games, social media, or any website.
- **"Santa hotline" or call-in recordings**, voicemails, or recordings of phone calls.
- **Stock or "royalty-free" site audio** (for example Pixabay, Uppbeat, Envato Elements). "Royalty-free"
  doesn't mean you can use it for AI training or redistribute it. For example:
  - Uppbeat's license says you may not use its content "to train AI models or for the purposes of machine learning."
  - Envato's Acceptable Use Policy bans using its assets "to train, develop, or enhance artificial
    intelligence models or machine learning algorithms."
  - Pixabay's terms ban data mining and scraping "for machine learning purposes," and they ban redistributing
    content in substantially unchanged form.
- **Another person's voice** without their written consent (see section 1).
- **Impersonations of real people**, including celebrities, actors, other well-known Santa performers,
  or any private person. Do your own Santa voice, not an imitation of someone else's.
- **AI-generated or voice-cloned audio** of any kind.

## 3. What you're agreeing to

By signing, you confirm or agree to each of the following:

1. **Your voice, your rights.** You're the speaker, or section 1's exception applies. You have the right to grant everything below.
2. **AI/TTS training.** We and others may use your recordings to train, test, and improve speech and TTS models.
3. **Public dataset.** Your recordings, their transcripts, and basic metadata (contributor ID, and your credit name if you choose one)
   may be published openly under the project's dataset license, **Creative Commons Attribution 4.0 (CC BY 4.0)**
   (see LICENSE-DATASET.md). That means anyone can copy, edit, and reuse them for any purpose, including commercially,
   as long as they give credit.
4. **Public model.** We may publicly release one or more voice models trained partly on your recordings. **The model's voice may sound like yours.**
5. **Anyone can use it.** Once released, the dataset and model can be used by anyone, anywhere, for any lawful purpose,
   including commercial products, ads, videos, and apps. We can't control or approve each use.
   Our model license asks users not to impersonate contributors or deceive people (see LICENSE-MODEL.md),
   but we can't guarantee everyone will follow it.
6. **Voice and publicity rights (separate from the dataset license).** CC BY 4.0 covers copyright only. Its section 2(b)(1)
   says publicity, privacy, and similar personality rights are *not* licensed. So this grant is separate: To the extent the law allows, you give permission for your voice in these recordings,
   and voices derived from it by the model, to be used as described above. You agree not to bring claims based on your
   voice, personality, or publicity rights against the project or people using the dataset or model within its license.
7. **No payment.** You're contributing voluntarily. You won't be paid and won't get royalties.
8. **How you're credited.** You ask that attribution under CC BY 4.0 be given as described in LICENSE-DATASET.md:
   one credit line plus a link to the project's contributors list, placed in the dataset or model card and docs.
   You agree that **this is reasonable attribution, and you don't require credit inside each generated audio clip.**
9. **No endorsement.** Publishing your recordings doesn't mean you endorse any product or use made with them.
   Nobody may present you as endorsing something because of them.

## 4. Withdrawing later

You can withdraw at any time by contacting [PROJECT CONTACT] with your contributor ID.

**What withdrawal does:**
- We'll remove your recordings from **future** releases of the dataset.
- We won't use them in **future** model training runs.
- We'll remove your credit name from future releases if you ask. Under CC BY 4.0 section 3(a)(3), you can also ask
  people reusing the dataset to remove your credit, to the extent reasonably practicable.

**What withdrawal can't do:**
- Dataset versions and models that were already released can't be recalled. CC BY 4.0 licenses are irrevocable,
  and other people may already have copies. Those copies can keep being used.

## 5. Credit

You choose one:
- **Credited:** we list the name or handle you give us in the project's contributors file. A pseudonym is fine.
  CC BY 4.0 allows attribution "by pseudonym if designated."
- **Anonymous:** we only use a random contributor ID (for example `C0042`). You'll still be covered by the group
  credit "OpenSanta contributors."

## 6. Your personal information

We keep your contact details, this signed form, and your spoken consent clip privately, only to verify consent and manage withdrawal requests.
- **Where your files go:** you upload through a Google Form, and Google stores the upload in the project's **private Google
  Drive**, which only project maintainers can access. Reviewed script clips later move to the public dataset. Your consent
  form, consent clip, and contact details never do.
- **Where consent records are kept:** your signed form, spoken consent clip, and contact details are moved into a
  **private, access-restricted Google Drive folder** in the project's own Google account. Only the project lead can access it,
  and anyone else needs the project lead's explicit approval. It's kept separate from the upload folder and from anything public.
- **Your Google account email:** uploading requires a Google sign-in, and the form records the email address of the account
  you use. We treat it as private contact information, the same as the rest of your details.

None of this private information is ever published in the dataset. Public files only show your contributor ID and, if you chose it, your credit name.

---

## Signature block

Copy this block, fill it in, and submit it with your recordings.

```
OPENSANTA PROJECT: CONTRIBUTOR CONSENT (v0.4)

Name or handle:            ______________________________
Contact (kept private):    ______________________________
Contributor ID (we assign): __________
Date (YYYY-MM-DD):         __________

Tick each box to confirm:
[ ] I am 18 or older.
[ ] I am the speaker in every recording I submit (or I hold all rights AND
    attach the speaker's written consent to every item below).
[ ] None of my audio is scraped, ripped, hotline/phone, stock-site, AI-generated,
    or an impersonation of a real person.
[ ] I consent to my recordings being used to train AI / TTS models.
[ ] I consent to my recordings being published publicly under CC BY 4.0.
[ ] I consent to the public release of a model that may produce a voice resembling mine.
[ ] I understand anyone may use the released dataset and model for any lawful purpose, including commercially.
[ ] I understand withdrawal only affects future releases and trainings, not versions already released.
[ ] I agree that credit via the project credit line + contributors list (in the dataset/model
    card and docs) is reasonable attribution; no per-clip credit is required.
[ ] I recorded the spoken consent clip myself, and I understand it may be compared to my
    recordings to confirm I'm the same speaker. It's kept private.
[ ] I am not paid and this is not an endorsement.

Credit preference (choose one):
[ ] Credit me as: ______________________________
[ ] Keep me anonymous (contributor ID only)

Signature (typed name is fine): ______________________________
```

---

## How we record consent (required: signed form + spoken consent clip)

Every contributor completes **both** steps below. We don't accept a submission that's missing either one.

### Step 1: Signed form
Fill in the signature block above and save it as `consent-<contributorID>.txt`.

### Step 2: Spoken consent clip (about 20 seconds)
Before your first script line, in the same session, setup, and microphone, record this as its own file,
`<contributorID>_consent.wav`. Read it in your **normal speaking voice, not your Santa voice:**

> "My name is [NAME OR HANDLE], contributor ID [ID]. Today is [DATE].
> This is my own voice, and I'm 18 or older.
> I've read the OpenSanta consent form, version 0.4, and I agree to it.
> I consent to my recordings being used for AI and text-to-speech training,
> and to their public release, along with a voice model that may sound like me, under the project terms."

The person reading this must be the speaker in the submitted recordings. If you're submitting someone else's
voice under section 1's exception, **that speaker** records this clip.

### How it all links together
- **Contributor ID.** We assign one ID per contributor. It appears in the signed form's file name,
  the consent clip's file name, and every audio file you submit.
- **SHA-256 manifest.** Before you submit, you generate `manifest-<contributorID>.sha256`, a list of the SHA-256 hash
  of each file, with `python tools/qc_check.py <folder> --pack <P#> --write-manifest` (see CONTRIBUTING.md).
  We check it again when we receive your files.
  It includes the hash of the consent clip and the signed form. This records exactly which files your consent covers.
  If a file is changed later, its hash won't match.
- **Same-speaker check.** We may compare your consent clip with your submitted recordings, by listening and/or with an
  automated speaker-verification tool, to confirm they're the same person. This is used only to verify consent.
  Results aren't published.

### Kept private
The signed form, the spoken consent clip, and your contact details are stored **privately** in the project's consent
records, with access limited to maintainers. They're **never included in the public dataset** and never used for training.
The public release may list each contributor ID with a "consent on file" flag and the hashes of the published audio files,
but nothing else from your consent records.
