AI Prompts for ElevenLabs Prompts for Ad Voiceovers

Cast the ad voice for [BRAND, PRODUCT, AUDIENCE]. Brand personality: [TRAITS: PREMIUM, PLAYFUL, TRUSTWORTHY, DISRUPTIVE]. Describe the target voice in production terms: gender or neutral, age impression, pace, warmth versus authority, and the read style (announcer, conversational friend, documentary gravitas). Then list what this brand’s voice must never sound like, and the audition criteria for ElevenLabs candidates.

I run ads on [CHANNELS: META / TIKTOK / YOUTUBE / RADIO / PODCAST]. Map the voice requirements per channel: TikTok wants native and unpolished, radio wants projection, podcast host-read style wants intimacy, YouTube pre-roll wants an instant hook. Tell me whether one ElevenLabs voice can flex across all my channels with different reads, or where I need a second voice, and define the read spec per channel.

Write a 30-second audition script to test ad voices for [PRODUCT]. Include: a pattern-interrupt opening line, one benefit statement, a price or offer with numbers, urgency without shouting, and the call to action. This script stress-tests everything an ad read needs. Give me the scoring sheet: attention, credibility, warmth, CTA authority.

Analyze the voiceover choices in my competitors’ ads: [DESCRIBE 3 COMPETITOR ADS: VOICE TYPE, STYLE, ENERGY]. Identify the category sound (what every brand in [CATEGORY] sounds like) and the differentiation opportunity: the voice character that would stand out in the same feed while staying credible for the product. Output the casting brief for that differentiated voice.

My product is in a trust-sensitive category: [CATEGORY: FINANCE / HEALTH / INSURANCE]. Define the voice and read style that builds credibility: the pace that signals confidence not sales pressure, where warmth belongs versus precision, and the delivery mistakes that trigger skepticism (over-enthusiasm, rushed disclaimers). Include how the required disclaimer section should be read: clear and unhurried, never mumbled.

Write my ad voiceover script for [PRODUCT, OFFER, AUDIENCE] at exactly [LENGTH: 15 / 30 / 60] seconds, budgeting roughly 2.5 words per second. Structure: hook (first 3 seconds), problem-agitate or desire, the offer with one number, CTA. Then output it formatted for ElevenLabs: punctuation placed for emphasis on [KEY PHRASE: THE OFFER / THE BRAND NAME], short sentences, no clause stacking.

Take my ad script ([PASTE]) and engineer the emphasis for ElevenLabs: identify the 3 words that carry the sale (usually the pain word, the promise, and the action verb), and reformat the text so the voice hits them: sentence position, punctuation before the key word, and one-word sentences where impact matters. Show before and after with a note per change.

Write 10 hook variants (first 5 seconds only) for my ad: [PRODUCT, AUDIENCE, CORE PAIN]. Vary the mechanism: question hook, bold claim, statistic, callout to the audience, pattern interrupt, mini-story. Format each for text-to-speech delivery. I will generate all 10 in ElevenLabs and cut them against the same body: hooks are the highest-leverage test in the account.

Format my required disclaimer for narration: [PASTE LEGAL TEXT]. Make it compliant but listenable: break into speakable sentences, spell out terms ElevenLabs might garble (APR, abbreviations, percentages), and set the read spec: measured pace, clear enunciation, no artificial rush. Flag anything in the legal text I should confirm can be reworded before changing it.

Convert my ad copy ([PASTE]) into a podcast-style host read for ElevenLabs: conversational, first person, sounds like a recommendation not a commercial. Include the natural-sounding transition in, a personal-experience framing ([ANGLE: HOW THE PRODUCT FITS A DAILY ROUTINE]), the offer code delivered memorably, and the casual sign-off. 60-90 seconds. The formatting should read like talking, not announcing.

Build my ad variant matrix: [NUMBER] hooks x [NUMBER] offers x [NUMBER] read styles (energetic, warm, authoritative). List every combination worth generating (not the full cube, the smart subset), the naming convention (hook2_offerA_warm), the generation order, and which single variable to isolate in the first test flight. Budget: [NUMBER] variants total.

For my locked ad script ([PASTE]), define 4 distinct read directions to generate in ElevenLabs: the settings and text-formatting adjustments that produce each (energetic launch read, calm premium read, urgent last-chance read, friendly recommendation read). Describe what each read signals to the buyer and which funnel stage or audience segment it fits.

Adapt my winning ad ([PASTE SCRIPT]) for [MARKETS / LANGUAGES]. For each: the translation approach that preserves selling rhythm over literal accuracy, the cultural adjustments (idioms, offer framing, formality level), ElevenLabs multilingual voice selection criteria to match my brand voice, and the local compliance items to verify. One ad, properly local everywhere.

Design the test flight for my [NUMBER] voiceover variants on [PLATFORM]: budget split, flight length for significance at my spend ([DAILY BUDGET]), the metric hierarchy (thumb-stop or hook rate first, then CTR, then CPA), the kill rule for losers, and the iteration protocol: what to generate next based on which read style wins. Output as a simple flight schedule.

Variant [ID] won my test: [PASTE ITS SCRIPT AND READ STYLE, PLUS RESULTS]. Autopsy why: hook mechanism, read energy, offer framing, or emphasis placement. Then productionize the insight: the template for future scripts, the locked read spec, and the 3 next variants that push the winning element further while the rest stays frozen.

My winning ad fatigues after [TIMEFRAME / FREQUENCY DATA]. Build the refresh system: the refresh cadence for [PLATFORM], what to rotate first (hook audio, not the whole ad), the ElevenLabs regeneration workflow that produces next-cycle variants in one session, and the fatigue signals that trigger early refresh: rising CPM, falling hook rate, comment tone shift.

Organize my ad audio library: naming convention across campaigns ([STRUCTURE: PRODUCT_HOOK_READ_VERSION]), the metadata to log per file (script, voice settings, performance), the evergreen assets worth keeping current (brand line read, CTA reads, disclaimer), and the quarterly prune rule. A searchable library means next campaign starts from proven parts, not from zero.

My 30-second ad won on [PLATFORM]. Adapt the voiceover for every other format: the 15-second cutdown (which sentences survive), the 6-second bumper (hook plus brand only), the 60-second podcast read (expanded with proof), and radio ([SPECS]). For each, the script edit and whether the same read style transfers or needs regeneration with different energy.

Multiple people now make ads for [BRAND]. Write the voice governance one-pager: the approved ElevenLabs voice and locked settings, the script formatting rules (emphasis, numbers, brand name treatment), the read specs per channel, QA sign-off before spend, and what requires approval to change. The brand should sound identical no matter who generated the file.

Run my quarterly ad audio review: performance by read style and hook type ([PASTE DATA]), cost saved versus studio production ([SESSIONS AVOIDED X TYPICAL COST]), where AI reads underperformed and a human session might be worth it, and the next quarter test agenda: the 3 hypotheses about voice, pacing, or format most worth testing. End with decisions, not observations.