Tutorials Video

Making a Synthetic Voice or Narration: Consent First

By Infinite Creation · Intermediate · ~13 min

A 1960s mid-century comic book style editorial illustration showing a narrator reading into a microphone in a booth, a loudspeaker producing matching sound waves beyond a glass window, and a stamped permission form on the table between them as a technician checks a level meter.

Overview

A synthetic voice is speech generated by software from text. Voice cloning goes one step further and imitates a specific person. Creators use both for narration, translation, placeholder tracks and accessibility. This guide is for anyone who plans to publish speech made by a tool. It puts the decisions that carry risk first: whose voice, with what permission, under what licence, and how you will tell your audience. It then covers what two makers state on their own pages. We did not test any tool on this page. Each statement about a tool comes from the maker's own page as read on 8 October 2026, and the Sources section says what each page established.

What You Need

The script, finished and proofread before you generate anything
A written answer to whose voice this will be
The tool's terms of use, read for commercial use and for voice rights
A plan for how you will tell listeners the voice is synthetic

Steps

Decide whose voice it is

There are three cases. A stock voice offered by the tool. A voice of yourself. A voice of another real person. The first needs only the tool's licence. The second needs your own informed decision. The third needs that person's written consent, a clear scope of use and payment terms where the work is professional. Never clone a voice you cannot prove you have permission to use.

Read the licence for the plan you will use

Commercial rights often depend on the paid plan. ElevenLabs' pricing page does not state commercial terms for its Free plan, and its Starter plan list begins with the Free features plus a commercial licence, which suggests the Free plan does not include one. We were unable to confirm that from the page. Read the terms for the exact plan before you publish.

Write the script for the ear

Write short sentences. Spell out numbers and abbreviations the way you want them spoken. Mark names with phonetic spellings if the tool mispronounces them. Generate a few lines first and listen before running the full script.

Generate in sections and listen to all of it

Generate a paragraph at a time so you redo a weak passage without redoing the rest. Listen to the whole result at normal speed. Listen for wrong stresses, odd pauses, mispronounced names and a flat delivery on emotional lines. A person must approve every minute that ships.

Label the result

Tell your audience the voice is synthetic, in the description or on screen, in plain words. The labelling page covers platform rules and audience trust. If the voice imitates a real person, say so and say they consented.

Keep a record

File the consent, the licence terms and the date you read them with the project. If a platform, a client or a lawyer asks later, the record answers the question.

Pro Tips

Use a stock voice when one fits. It removes the consent question.
Do not use a cloned voice for deceptive purposes, impersonation or fake endorsements. That harms people and breaks platform rules.
A person reading the script usually sounds better on emotional material. Use synthetic voice where flat delivery is acceptable.

Knowledge Base

What You'll Learn

The technology is routine now. The deciding issues are consent, licensing and honesty with the audience.

What Respeecher says about itself

Respeecher's page describes enterprise voice cloning and speech synthesis, a real-time text to speech API, a voice library and a Pro Tools plug-in. It lists film and television, animation, games, broadcasting, sports, podcasts, audiobooks, music, advertising, e-learning and healthcare among its uses. It describes strict consent mechanisms and voice origin tracking as ethical protocols. Those statements are the company's own and were not independently checked. No 2026 studio deal reporting was found for it.

Why professional work asks for more

Film and television productions use contracts, performer consent and payment for any voice recreated by software. The union's own summary of the tentative 2026 SAG-AFTRA TV/Theatrical agreement covers replicas and consent for dubbing, and our guide to it lists what it says. It is a summary of a tentative agreement, not the contract, and other contracts are not covered. If your work is professional, read the voice ethics page and take advice on the contracts that apply to your production.

Where This Fits

A generated voice replaces the recording stage, and Best Podcast Setup for Beginners covers the microphone, room and editing work that a spoken-word production otherwise starts with.

FAQ

Q: Am I allowed to clone my own voice?
A: Yes, the tools offer it. Read the terms for who owns the cloned voice model and where your recordings are stored.

Q: Do I have to label synthetic voices?
A: Platforms and some laws increasingly require it. The labelling page covers current practice. We have not read the primary legal texts, so check the rules where you publish.

Q: Is a synthetic narration as good as a recorded one?
A: We did not test any tool. An RNIB study of synthetic voices for audio description, which we know only from secondary reports and did not read, found them acceptable for factual and documentary material and weak on emotion in drama.

Vocabulary

Voice cloning Training or adapting a model to imitate a specific person's voice from recordings. It raises consent and impersonation issues that a stock voice does not, so the person whose voice is copied should agree in writing first.

Text to speech Software that turns written text into spoken audio. Modern systems use neural networks and produce natural pacing, but still misjudge stress, names and emotion, so a person should listen to the full result before it ships.

Informed consent Permission given by a person who has been told exactly how their voice or likeness will be used, for how long, and for what payment. Vague permission does not count.

Sources

Two maker pages were read on 8 October 2026. No voice was generated for this guide. Checked 8 October 2026.

ElevenLabs, pricing. Lists plan features. Does not state commercial terms for the Free plan. The Starter list begins with everything in Free plus a commercial licence.
Respeecher. Describes enterprise voice cloning, a real-time text to speech API, a Pro Tools plug-in, its list of industries, and its own consent and voice origin statements.

The union summary covered in the linked replicas guide was read, not the full contract, and no law was read for this page. Statements about labelling rules and about broadcaster research come from our site pages and trade coverage, and were not read in their primary form. The ElevenLabs commercial terms for the Free plan remain unconfirmed.