How to Use ElevenLabs to Publish a Newsletter's Audio Edition

The short answer: a weekly newsletter of 8,000 characters spends 32,000 ElevenLabs credits a month at 1 credit per character, just over Starter's 30,000, and the embeddable player that puts it on your site, Audio Native, starts on Creator anyway, at $22 a month with 121,000 credits (pricing page, Audio Native plans, checked on 26 October 2026). So budget for Creator. The sequence: get the founder's written consent, record two clean minutes, create an Instant Voice Clone, build the issue in Studio, open it with a spoken AI disclosure, let Studio Agent add the bed and the breaks, have the founder approve the export, and publish it through Audio Native.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

Before you switch anything on:

1. Count characters, not words. Your editor shows the character count. For example, 4 issues x 8,000 characters = 32,000 credits, about a quarter of Creator's 121,000, which leaves room for blog posts and retakes (pricing FAQ).
2. Start from a recording you already own. A clean podcast intro or keynote clip from the founder can serve as the clone sample, if it has one speaker, no music and no room echo.
3. Pick the measure now. Audio Native has a listener dashboard (Audio Native guide). Note the issue's current click-through rate and watch both for six issues.

What the plans include and cost is in the ElevenLabs glossary entry.

What has to be in place

• Written consent from the founder. ElevenLabs' Prohibited Use Policy forbids cloning any living person's voice without explicit and informed consent. Write down what the clone may read (the newsletter and blog), where it runs, for how long, and how the founder withdraws.
• A Creator subscription, for Audio Native and Professional Voice Cloning later if you want it.
• Access to your site's HTML or a CMS that takes an embed, such as WordPress, Ghost, Webflow or Framer, each of which has its own guide.

Seven steps from draft to player

1. Record the sample properly

The Instant Voice Cloning guide asks for 1 to 2 minutes of clean audio with no reverb or background noise, warns that more than 3 minutes adds little and can hurt, and puts the target loudness at -23 to -18 dB RMS with a true peak of -3 dB. The clone copies the delivery too, so have the founder read a past issue in the tone the newsletter should have, not a flat test script. Export MP3 at 192 kbps or above.

2. Create the clone, with the consent on file

In the dashboard open Voices, click the plus icon, choose Instant Voice Clone, upload the sample, name it, confirm you have the right and consent to clone the voice, and click Save voice (cloning guide). Store the signed consent next to the voice name.

3. Build the issue in Studio

Open Studio, select Upload and add the issue as a text file; imported documents get their chapters detected automatically (Studio guide). Assign the clone in the Voiceover panel, then use the Pronunciations editor for product names and acronyms, with alias or phoneme rules.

4. Say it is AI in the first sentence

Article 50(4) of the EU AI Act, applicable from 2 August 2026, requires deployers of AI that generates audio constituting a deep fake to disclose that it is artificially generated. Article 3(60) defines a deep fake as content that resembles existing persons and would falsely appear authentic, and Article 50(5) asks for the disclosure at the latest at first exposure (consolidated AI Act). A founder's cloned voice reading the founder's words fits that description. So the first paragraph in Studio is: "This audio edition is read by an AI clone of [name]'s voice, made with [name]'s consent." Put the same line in the player title or under it. ElevenLabs' own policy adds that you must not remove its watermarks or provenance signals (policy).

5. Let Studio Agent dress it, and the founder sign it off

Studio Agent is the editor's built-in agent, on all plans: describe what you want and it builds on the timeline, adding clips, voiceovers, sound effects and captions (Studio guide). Ask it for a soft sound effect under the disclosure line and a short sting between sections. Its chat is billed in credits by tokens, so keep requests specific.

The human check: the founder listens to the full export before anything is published, and marks every line that sounds unlike them. When a paragraph's button reads Regenerate, you can take two new versions without spending credits, as long as the voice and text stay the same (Studio guide), so fixing a flat sentence is cheap.

6. Export, then publish to Audio Native

Select Export to render the project (Studio guide). Under Audio Tools, open Audio Native, set the player colours, add your newsletter archive or whole domain to the allowlist, and copy the embed code into the page (Audio Native guide). A player embedded on a page outside the allowlist will not work.

7. Update without breaking the player

When an issue changes after publishing, open it from the Audio Native page with Edit Audio, change it in Studio, export as MP3 (required even if you do not need the file), then Export, Publish, Audio Native, and select the new published version (update article).

Warnings in ElevenLabs' docs

• Longer samples are not better samples. Past 3 minutes the clone gains little and can get worse (cloning guide).
• Uploaded audio cannot go to distribution platforms. Studio says so of files you upload into a project (Studio guide); generate the narration in ElevenLabs.
• Downgrading locks voices. A plan with fewer voice slots hides the clones above its limit until you upgrade (downgrade article).

The checkbox, and the free trial run

The clone dialog asks you to confirm consent with a tick, and the policy states that complying with the laws that apply to you is your responsibility (policy). The tick records that you claimed consent; the proof that you had it is yours to keep, which is why the signed note in step 2 matters more than the button.

Before any of that, run one issue on the free plan. Its 10,000 monthly credits (pricing page) cover one 8,000-character issue in a stock voice from the library, with no clone and no consent form. Post it as a plain MP3 link in the next issue and count the plays. If nobody listens to a stock voice, a cloned one will not change much; if they do, the clone is the upgrade.

FAQ

Does the founder have to record anything after the clone exists?

No. The clone reads any text you give it. The founder's job becomes approving each export, which takes as long as the issue runs.

Does the disclosure line apply outside the EU?

Article 50 is EU law. For listeners elsewhere, ElevenLabs' own policy still requires you to disclose when people are interacting with AI where the law requires it or where not doing so would materially mislead them (policy). One line at the start covers both.

Can the founder withdraw consent later?

The consent note should say so and how. ElevenLabs lets a voice owner report unauthorised use of their voice (policy); in practice, delete the clone and replace published audio from the list of issues it voiced.