AI Prompts for ElevenLabs Prompts for Voice Cloning

I want to clone my voice with ElevenLabs. My content: [WHAT I MAKE: VIDEOS, PODCAST, COURSES, CLIENT WORK]. Audit my use cases: which tasks a clone genuinely serves (narrating scripts I wrote, updating old content, scaling formats I cannot record for), which it serves badly (spontaneous conversation, interviews), and the hours per month it realistically saves. Verdict: instant clone or professional clone, and why.

Walk me through the rights checklist before cloning: I am cloning [WHOSE VOICE: MY OWN / A TEAM MEMBER / A CLIENT WITH PERMISSION]. For anyone other than myself: the documented consent I need, what the agreement should cover (use scope, revocation, ownership if we part ways), and the uses that stay off-limits regardless. Also: ElevenLabs verification requirements and my own policy for where my cloned voice may appear.

Plan my clone training recording: the environment (room treatment on a budget, mic distance, noise floor), the sample length and variety ElevenLabs needs for [CLONE TYPE: INSTANT / PROFESSIONAL], and the read content: natural conversational material in my real delivery style, not stiff script reading, since the clone learns my habits. Write the session checklist and the 3 sample scripts to read, matched to how I actually talk in my content.

My voice changes by context: [DESCRIBE: ENERGETIC ON CAMERA, CALM ON PODCASTS, FORMAL FOR CLIENTS]. Decide my cloning strategy: one clone from my most-used delivery style, or multiple clones per context. Weigh: which style covers the most content hours, whether settings adjustments can stretch one clone across contexts, and the maintenance cost of multiple clones. Recommend and specify the training sample style for each clone I should make.

Define the acceptance test for my clone before I build workflows on it: the 4 test passages (normal narration, names and numbers, an excited passage, a quiet sincere passage), the blind test protocol (mix clone and real recordings, have [PERSON WHO KNOWS MY VOICE] identify which is which), and the pass bar. A clone that fails with people who know me needs a better training sample, not better settings.

My clone sounds [PROBLEM: FLATTER THAN ME / TOO FAST / MISSING MY WARMTH / WRONG ON CERTAIN SOUNDS]. Diagnose systematically: is the gap in the training sample (energy, variety, audio quality), the settings (stability too high flattens personality), or the script formatting (my real speech has rhythms the text does not encode)? Give me the isolation test for each cause and the fix in order of likelihood.

Calibrate my clone settings: generate this test paragraph ([PASTE A PARAGRAPH IN MY NATURAL STYLE]) at [3-4 SETTING COMBINATIONS] and define what to listen for in each: stability versus personality trade-off, similarity boost effects, style exaggeration risks. My content is [TYPE], so define the winning criteria: which version would my audience accept as a normal recording of me?

Ironically, I now write scripts for my own cloned voice. Analyze my natural speech patterns from this transcript of me talking ([PASTE A REAL TRANSCRIPT]): my sentence lengths, filler habits worth keeping (the natural ones), pet phrases, how I emphasize. Write the style guide for scripting my clone so generated content keeps my verbal fingerprint, and rewrite this sample script ([PASTE]) to match it.

My clone mispronounces things I say distinctively: [LIST: NAMES I SAY OFTEN, REGIONAL PRONUNCIATIONS, BRAND TERMS]. Build my pronunciation fingerprint file: each term, how I actually say it, the respelling that reproduces my version. My audience knows how I say these; the clone saying them differently is the tell.

Map what my clone can and cannot do: test it across the emotional range my content needs ([LIST: ENTHUSIASM, SERIOUSNESS, HUMOR, EMPATHY]) using one test line per emotion ([PASTE LINES FROM MY REAL CONTENT]). Log where it convinces and where it falls flat. Output my clone usage policy: which content types get the clone, which still get the real microphone.

Build my weekly pipeline using my voice clone: content types ([LIST: VIDEO NARRATION / PODCAST SEGMENTS / COURSE UPDATES / SOCIAL AUDIO]), the script-to-published flow per type, where the clone slots in versus where I record live, the QA pass that catches un-me moments before publishing, and the time math: hours saved per week versus my old record-everything workflow.

I have [NUMBER] old pieces of content with outdated audio or no audio: [DESCRIBE: OLD POSTS TO VOICE, VIDEOS WITH DATED INFO, COURSES NEEDING UPDATES]. Plan the revival: which backlog items gain most from my cloned voice (evergreen content first), the update workflow (regenerate the dated sentences, keep the rest), and the production order by traffic potential. Estimate sessions saved versus re-recording it all.

ElevenLabs can render my cloned voice in other languages. Plan the multilingual strategy: which languages my audience data supports ([PASTE ANALYTICS / MARKETS]), the script translation workflow that keeps my persona (my humor and directness must survive translation), the native-speaker review step before publishing, and the pilot: one high-performing piece in [LANGUAGE], measured against a defined success bar.

Design my content continuity plan: when I am [SITUATION: TRAVELING, SICK, ON LEAVE], the clone keeps essential content flowing. Define: which content types are pre-approved for clone production in my absence, who operates it and under what rules (scripts approved by [WHOM], no new topics), the disclosure approach if any, and the hard limits: things my voice never says without me, listed explicitly.

I produce content for [CLIENTS / MY TEAM] and want to offer voice cloning as a service: cloned client voices for their content. Build the service framework: the consent and contract template essentials (scope, revocation, data handling), pricing structure ([MODELS: PER PROJECT / RETAINER]), the recording session I run for their training sample, and the governance: their voice is only ever used with their written sign-off per project.

Audit my voice clone security: who has access to my ElevenLabs account (and should), whether my training samples are stored anywhere insecure, what public audio of me exists that others could clone without consent, and my monitoring approach: how I would find out if my voice appeared somewhere I did not put it. Output the hardening checklist, ranked by risk.

Write my voice clone disclosure policy: for each content type I produce ([LIST]), whether disclosure is required by the platform, legally advisable, or a trust decision. Consider my audience relationship ([DESCRIBE]) and what discovery-without-disclosure would cost me. Output: the policy per content type and the standard disclosure wording where I choose to use it.

Help me draw the line my clone never crosses: content where my real voice is the point (personal stories, apologies, sensitive announcements, anything emotional my audience deserves live), versus content where the clone is a legitimate production tool (narration, updates, translations). Write the boundary as a one-page policy with the reasoning, so future-me under deadline pressure does not blur it.

Set up drift monitoring for my clone: the quarterly check (regenerate my benchmark passages, compare against the originals and against my current real voice, since voices age and styles evolve), the signals that mean retraining (audience comments, my own ear, accumulating fixes), and the retraining protocol: new sample session, re-run the acceptance test from my quality benchmark.

Run my annual voice clone review: hours saved this year ([ESTIMATE FROM PIPELINE DATA]), content produced with the clone versus live, any incidents or near-misses (misuse, un-me moments published, disclosure questions), whether my usage policy boundaries held, and next year’s decisions: expand use cases, retrain, add languages, or pull back anywhere. Output: what worked, what to change, the updated policy.