Project P10 — Your Voice, With Consent

“…the sheep follow him, for they know his voice. A stranger they will not follow, but they will flee from him, for they do not know the voice of strangers.” — John 10:4–5 (ESV)

Chapter: 10 — Creating with AI: Voice and Audio Due: End of Week 10 Submit: A link to a shared folder or document (Google Drive, OneDrive, or a single PDF portfolio) containing your completed consent form, your audio clip (file or link), your ethics memo, and screenshots. Not GitHub — see Submission. Lab and account setup is in Appendix A. Allowed tools: For the writing — the script, the memos, your reflections — BoodleBox is your default home in this edition (sign in with your Concordia account at box.boodle.ai, start a chat, and @-mention a strong model to draft with). Because BoodleBox is text- and image-first as of 2026, the voice generation itself happens in a consumer voice/audio tool reachable from a browser or phone — a TTS/voice-cloning tool (ElevenLabs or similar, on its free or trial tier where available; instant cloning is often a paid feature, and the zero-budget path is the documented TTS stand-in in the Normal tier), and, for the Medium tier, Google NotebookLM. BoodleBox may help you draft the memo — but the judgment in it must be yours; if you cannot reach it, any free assistant drafts the same. See the tool directory in Appendix B. AI policy: You may use AI to generate the voice and to draft prose. You may not delegate the judgment — consent, disclosure, and the misuse-boundary decisions are yours, and the memos are graded on exactly that. Every deliverable that used AI carries an honest one-line AI usage note. You choose the tool; you own the verdict.

The Setup

You are Dana Whitfield, owner of Whitfield Learning — a two-person continuing-education business that produces short audio lessons, course welcome messages, and phone-tree greetings for local employers. You narrate everything yourself, and it has become the bottleneck: every time a course detail changes, you re-record; every edit means booking quiet time and doing another take. Your voice is your brand, and re-recording it is eating your week.

A colleague suggests you clone your own voice so you can produce narration by typing instead of re-recording — change the script, regenerate, done. It is a genuinely good idea, and you are going to test it.

But the moment you start, the harder questions arrive. If listeners hear “Dana” narrating a lesson you never actually spoke, do you owe them a heads-up? A well-known local career coach has a distinctive voice, and cloning her for a promo would be devastatingly effective — is that a line, or a temptation you should name and refuse? And if your cloned voice leaks, what stops someone from making “Dana” say things Dana never would? You are a small business built on trust. You will not spend that trust carelessly on a tool you have not thought through.

So this week you will do three things: clone (or high-fidelity narrate) your own voice under a signed consent form, produce a short disclosed clip from a provided script, and write the ethics memo that turns “this is cool” into “this is how we will and will not use it.”

Setup (the starter)

This chapter ships two starter artifacts in code/. Use them directly.

• code/voice-consent-form.txt — a plain-language synthetic-voice consent and authorization template: whose voice, permitted uses, forbidden uses, disclosure, retention, and revocation. You will complete it for your own voice.
• code/voice-script.txt — the short script you will render in your voice, including a spoken disclosure line to model good practice.

Copy them into your working folder and make them yours — the consent form especially is a document you fill in, not a form you just attach. In BoodleBox you can attach the script as Knowledge (the “Attach Knowledge” control, as of 2026) so a model drafts your memo grounded in the exact clip you are producing — but the consent form’s judgment is yours to supply, not the model’s.

Learning Targets

You will demonstrate that you can:

• Treat your own voice as an asset governed by consent — completing a scoped, revocable authorization, not shrugging past it.
• Produce disclosed synthetic audio with a consumer tool, from your own voice, on a real script.
• Reason clearly about disclosure — telling listeners when a voice is synthetic, and why the small awkwardness is worth it.
• Draw misuse boundaries — including the out-of-band verification defense against voice-cloning fraud — and defend where the line falls.
• Make a judgment call an AI cannot make for you: whether and how a real organization should adopt voice cloning at all.

Normal Tier

Goal: Clone or high-fidelity narrate your own voice under a signed consent form, produce one short disclosed clip from the provided script, and write a one-page ethics memo.

Required features

1. Completed consent form. Fill in code/voice-consent-form.txt for your own voice. Every field: whose voice, the specific permitted uses, the explicit forbidden uses, the disclosure standard, the retention/deletion plan, and how consent can be revoked. Sign and date it (a typed name is fine).
2. A voice, honestly sourced. Create an instant clone of your own voice with a consumer tool’s free or trial tier where available (ElevenLabs or similar — instant cloning is often a paid feature, so check Appendix B and the vendor’s current pricing page before you count on it). If no cloning tool is available in your environment — including if cloning sits behind a paywall — use a high-quality TTS voice as a documented stand-in and say so plainly in your memo; the stand-in path costs nothing and fully satisfies this project. Do not clone anyone else’s voice as a substitute.
3. A short disclosed clip (~30–60 seconds). Render the provided script code/voice-script.txt in your voice, including the spoken disclosure line at the end. Export the audio (or a shareable link to it).
4. Disclosure, present and clear. The clip must identify itself as AI-generated — the spoken line in the script, plus a written disclosure in your portfolio. A listener must be able to tell this voice was synthesized.
5. A one-page ethics memo (~400–600 words) — draft it in BoodleBox if you like (open a chat, @-mention a strong model, and have it help you structure and tighten the three sections), then make the judgment yours — covering three things explicitly, each under its own heading:
   • Consent — whose voice, on what scope, revocable how.
   • Disclosure — when and how you will tell listeners a voice is synthetic, and why.
   • Misuse boundaries — what you will never do with a cloned voice (including cloning others without consent), and how you will guard your own clone from misuse (including an out-of-band verification note — code word / call-back / dual approval).
6. Portfolio + honest AI-usage note. Assemble the consent form, the audio (file or link), the memo, and screenshots of your tool showing the generation, into one shared folder or PDF. Include a one-line AI usage note: what you used, what it got wrong, what you verified.

Normal-tier rubric (out of 100)

Medium Tier (+up to 25% extra credit)

Bring in the second half of the chapter: NotebookLM Audio Overviews.

1. Make an audio overview. Gather 3–5 documents on one topic — a real set for Whitfield Learning (course materials, a syllabus, articles) or your own work. Nothing confidential. Upload them to NotebookLM and generate an Audio Overview.
2. Fact-check one claim. Pick one specific factual claim the two hosts make — a number, a date, a “the source says.” Open the actual source and find (or fail to find) the sentence that backs it. Record the result: exact / approximate / unsupported.
3. Write it up (½ page). What the overview got right, where it oversimplified or over-agreed, the caps/format limits you hit, and one thing you would not act on from the audio alone. Include screenshots and an AI usage note.

The grade is in the verification, not the fun: an audio overview drafts your understanding; you own the verdict on what it means.

Hard Tier (+up to 25% additional extra credit)

This is the judgment no AI can make for you, and it is graded as such.

Write voice-cloning-decision-memo (one page, addressed to “Dana Whitfield / the Whitfield Learning owners”) that makes and defends a single recommendation: Should Whitfield Learning adopt voice cloning as a standard practice — and if so, under exactly what rules? You must:

• Draw the line. List the specific uses you will permit (e.g., re-narrating your own course updates) and the specific uses you will forbid (e.g., cloning the local career coach’s voice for a promo; using your clone for anything without disclosure). A memo that says “use it for everything” or “never touch it” fails — the grade is in the discrimination.
• Name the failure you are accepting and how you will contain it — what happens if your clone leaks, and what out-of-band control (code word / call-back / dual approval) protects Whitfield’s clients from a “Dana” impersonation.
• Tie it to the spine rule — you choose the tool; you own the verdict — and to at least one concrete control from your completed consent form (scope, disclosure standard, revocation, retention).

A machine can generate the voice in seconds. It cannot decide whether Whitfield Learning should clone a voice, where the honest line falls, or who answers when it goes wrong. That decision is the deliverable.

Submission

Submit a link to a shared folder or a single PDF portfolio (Google Drive / OneDrive / PDF) — not GitHub; you are a professional, not a programmer, and these are documents, not code. Include:

Project-P10/
  consent-form.pdf            (or .txt/.docx — your completed voice-consent-form)
  voice-clip.mp3              (or a shareable link to the audio)
  ethics-memo.pdf             (or .docx — the one-page memo)
  screenshots/                (tool showing the generation; disclosure visible)
  audio-overview-writeup.pdf  (Medium tier only)
  voice-cloning-decision-memo.pdf  (Hard tier only)

Set the share link so your instructor can open it. Use placeholders for anything private, and never upload anyone else’s voice or any confidential material. Submission mechanics are in Appendix A; the consent/disclosure standards are in Appendix C.

Hints (Read Before You Begin)

• Fill the consent form first. It forces you to decide the scope before you generate anything. If you clone first and think about consent later, you have done the week backwards.
• Use your own voice, or a clearly labeled TTS stand-in. There is no honest shortcut here. Cloning a friend’s or a celebrity’s voice “just for the assignment” fails the assignment on its own terms.
• Do not skip the spoken disclosure line. It is in the script for a reason — modeling honest practice is the point, not an afterthought.
• The memo is not a summary of the chapter. It is your decision for this business. Specifics beat platitudes: name the local-coach temptation and refuse it out loud.
• Verify the NotebookLM claim (Medium). “The hosts sounded sure” is not verification. Find the sentence in the source or report that you couldn’t.
• AI may draft; you must decide. If a chatbot writes your memo’s judgment, you have missed the entire assignment. Draft with it if you like; then make it yours and own it.

What Mastery Looks Like (Beyond the Rubric)

A mastered submission is one where I can read your consent form and your memos and tell that you are in the loop. Your consent form has a real, specific “forbidden uses” list, not a shrug. Your disclosure is present and unembarrassed. Your misuse boundaries name the out-of-band defense and mean it. And your Hard-tier memo does the hard thing — it draws a line between a permitted use and a forbidden one and defends the line, including the temptation it refused. The clip is table stakes; the consent, the disclosure, and the judgment are the project.

Coach’s Note — The temptation this week is to spend all your time perfecting the audio and treat the memo as paperwork. That is exactly backwards. Anyone can make a good-sounding clip now — that is the whole problem the chapter is about. What almost no one does well is decide, in advance and in writing, where the honest line falls. Spend your best hour on the memo. That is the hour that would have saved Dana’s business.

When You’re Done (a short checklist)

• Consent form is completed for your own voice, with real forbidden-uses and revocation entries.
• The clip renders the provided script and ends with the spoken disclosure line.
• Disclosure is present both spoken and written; a listener can tell it is synthetic.
• The ethics memo has all three sections: consent, disclosure, misuse boundaries (with an out-of-band note).
• You cloned/narrated only your own voice — no one else’s.
• (Medium) Audio overview made and one claim fact-checked against the source.
• (Hard) Decision memo draws and defends a permit/forbid line.
• Portfolio is shared as a link, with screenshots and an honest AI-usage note.

A theological footnote. The sheep know the shepherd’s voice, and they flee a stranger’s, because a voice is the mark of a person — a bearer of trust that a copy cannot carry (John 10:4–5, ESV). This project is the smallest possible rehearsal of honoring that. Your voice is yours to give, on terms you set — that is the consent form. When you send it out to speak words you did not personally record, you owe the listener the truth about what they are hearing — that is the disclosure. And you never lend your neighbor’s voice to a lie, nor leave your own where it can be turned into one — that is the misuse boundary, and it is the Eighth Commandment in a microphone age. The tool can copy the sound in seconds. It cannot supply the honesty. That part was always going to be yours.

See you next week.