AI voice cloning for teachers: consent, quality and cost

Why use a teacher’s own voice, and when a stock voice is fine

Teachers and course creators ask about an AI voice clone for one of two reasons. Students already know the teacher from live classes or recorded videos, and a strange voice in the AI lessons breaks that thread. Or the course is sold on one person’s name, and their voice is part of the product. In both cases, the clone buys familiarity and consistency: the lesson sounds like the person the student signed up to learn from, every time.

Do not expect a clone to teach better on its own. In a 2017 randomized study in Computers & Education, Craig and Schroeder compared a virtual human speaking with an older text-to-speech engine, a modern one, and a recorded human voice. The modern engine was rated at the same level as the human voice for facilitating learning and for credibility, and the authors concluded that their results “call into question” the older claim that human voices teach better. The voice needs to be clear and natural. It does not need to be yours.

So a stock voice is often the right choice:

• Corporate, compliance or product training, where no single teacher is the face of the course.
• Programs with many instructors, where one cloned voice would misrepresent who wrote the material.
• A teacher who would rather not be cloned. That answer should end the discussion, with no penalty.
• Pilots, before you know whether students will use the AI lessons at all.

How much audio does an AI voice clone need?

Less than most people expect, but quality matters more than length. Requirements vary by tool and by quality tier, so check the vendor’s own documentation:

• Tutezy clones a teacher’s voice from a 15-second sample, recorded or uploaded by the teacher after they accept a consent statement in the product.
• ElevenLabs recommends about one to two minutes of clean audio for its Instant Voice Cloning, and for Professional Voice Cloning at least 30 minutes, ideally closer to two to three hours. Its documentation says you can only create a Professional Voice Clone of your own voice, and you are asked to verify it.

Short samples capture what your voice sounds like. Longer ones capture more of how you speak: rhythm, emphasis, range. ElevenLabs puts the rule simply, “good consistent input = good consistent output,” and the same holds for any tool. Recording tips that apply everywhere:

• Quiet, soft room. No fans, traffic or echo. ElevenLabs suggests a thick duvet or quilt to dampen a room if you have no treated space.
• Same distance from the mic for the whole take; ElevenLabs suggests about two fists away. Moving closer and farther changes the tone the clone learns.
• Your natural teaching pace and energy. The clone copies the performance you give it, so do not read like a newsreader if you teach like a coach.
• Varied sentences: a statement, a question, a number, an instruction, an encouraging line. Monotone input produces a monotone clone.
• One speaker, no music, no effects. Record a few takes and keep the cleanest, checked on headphones.

How to get voice cloning consent from a teacher

A checkbox in a sign-up flow is not enough when the voice belongs to an employee or contractor. Put the terms in writing, separate from the employment contract, and keep the signed record with the voice. Cover these eight points:

1. 1

   The right person consents

   The person whose voice it is, in writing, from a sample they record or upload themselves. Never from found audio, old lecture recordings or a colleague’s clips.
2. 2

   Scope

   Which courses, subjects and languages. Lessons only, unless they agree otherwise. Name the excluded uses: advertising, phone calls, social media, anything they have not reviewed.
3. 3

   Duration

   A fixed term, such as one academic year or the life of a named course, with renewal by fresh agreement rather than silence.
4. 4

   Withdrawal and deletion

   How to withdraw, how fast the voice is disabled, what is deleted (the voice model and the sample), and written confirmation when it is done.
5. 5

   When they leave

   Decide now. A fair default: disabled on the last day and deleted, with lessons switching to a stock voice, unless they sign a separate paid license.
6. 6

   Compensation

   State it, even if the answer is that it is part of the role. For creators licensing a voice to a platform or school, agree a fee or share.
7. 7

   Minors

   Do not clone students’ voices. If a minor’s voice is ever involved, get a parent or guardian’s written consent first.
8. 8

   A named contact

   Who the teacher contacts to review uses, report misuse or withdraw, and how quickly that person must respond.

What to tell students

Tell them plainly, in two places: the course page and the opening of the first lesson. One sentence is enough: “Lessons in this course are taught by an AI tutor that speaks in Ms. Rivera’s voice, with her permission. She reviews the lessons, but she is not speaking live.” Add a line to your FAQ for parents if your learners are young.

Disclosure costs nothing and prevents the worst outcome, a student or parent discovering later that the “teacher” was not the teacher. It also marks a line you should not cross: do not use the cloned voice for phone calls or voicemails to students or parents. In a February 2024 declaratory ruling, the FCC confirmed that AI-generated voices, including voice cloning, count as “artificial” voices under the Telephone Consumer Protection Act, so such calls generally need the called party’s prior express consent.

What US law says about cloning a voice

Voice cloning law in the US is a patchwork that is still being written. As of September 2026, the main markers are:

• Tennessee: the ELVIS Act. The Ensuring Likeness, Voice, and Image Security Act of 2024 was signed on March 21, 2024 and took effect July 1, 2024. It adds a person’s voice to the property rights already protected in their name, photograph and likeness. It creates liability for publishing or making available someone’s voice with knowledge that the use was not authorized, requires a parent’s or guardian’s authorization for minors, and reaches tools whose primary purpose is producing a particular person’s voice without authorization.
• California: AB 2602. Signed September 17, 2024, it added Section 927 to the Labor Code. For new performances fixed on or after January 1, 2025, a contract clause allowing a digital replica of a person’s voice or likeness in place of work they would otherwise have done in person is unenforceable if it lacks “a reasonably specific description of the intended uses” and the person was not represented by a lawyer or a union in the negotiation. For a school or company employing teachers in California, that is a strong reason to write the specific, separate consent described above.
• The FTC. The FTC’s Impersonation Rule, in effect since April 1, 2024, covers impersonating government agencies and businesses, and the FTC has proposed extending it to impersonating individuals. In its 2024 write-up on AI-enabled voice cloning, the agency said plainly that “there is no AI exemption from the laws on the books.”
• Congress. The NO FAKES Act of 2026 (S. 4591), a bill “to protect intellectual property rights in the voice and visual likeness of individuals,” was ordered reported by the Senate Judiciary Committee in June 2026. It is not law.

These are examples, not a complete survey; check the rules in the states where your teachers and students are. This section is general information, not legal advice.

Keeping a cloned voice secure

A voice clone is a credential: anyone who can select it can make the teacher appear to say things. Before you create one with any vendor, get answers to five questions: who else can select the voice, whether it can be exported, whether the sample is used to train anything else, how disabling works, and how deletion is confirmed. The AI tutoring software checklist covers the wider data questions.

As an example of the answers to expect, Tutezy’s security and consent page commits to the following. A voice is cloned only from a sample recorded or uploaded by the teacher or account holder, after they accept a consent statement, and the consent record is kept with the voice. Public figures and colleagues from found audio are refused. The voice is private to the institute: nobody else on the platform can pick it. Disabling it in settings switches lessons to a stock voice immediately, and on request deletion removes both the voice model and the sample. Customer content is not used to train models.

Stock voice, cloned voice or human recording?

You can mix all three: the teacher’s real recording for the welcome video, a cloned voice for the interactive lessons, and a stock voice for courses where no single teacher owns the material. For how interactive lessons fit a course you already sell, see How to add an AI tutor to your online course.

What a teacher voice clone costs

Before you compare prices, check how each vendor charges for cloning (per voice, as part of a subscription, or by usage), how long a voice stays available if you stop paying, and whether deletion is included.

With Tutezy, a teacher voice clone is $20 one-time per voice (₹2,000 in India), usable across every course in the institute, in English and Hindi. Stock voices, including American English, are free. Lesson minutes cost the same $0.03 per learner-minute whichever voice teaches. So an independent course creator pays $20 once; a tutoring company that clones six tutors pays $120 once; and 100 students taking a 30-minute lesson costs $90 in either case. The full rate card is on the pricing page, and the business case for tutoring companies is in How to scale a tutoring business with AI.