ElevenLabs prompts for audiobook pacing and character voices
Paste-ready narration briefs for chapter pace, character voices, dialogue, inner life, action and nonfiction tone. Fill [CHAPTER TEXT], [VOICE AGE] and [PACE]. Original voices only, not famous narrators.
In short: This page contains 30 copy-paste ready prompts, organized into 6 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Drafted with AI writing assistance and checked against our editorial policy.
Chapter Pacing
5 promptsOpening chapter slow burn
1/30✨ What it does
Directs a slow first-chapter read with room after paragraphs and a single pause before dialogue. Keep the take if the opening still feels like sitting down, not a trailer.
ElevenLabs audiobook take for the first pages of a novel. Voice: a [VOICE AGE] narrator, [ACCENT], warm chest tone, never a celebrity impersonation. Pace: [PACE], unhurried, as if the listener has just sat down. Read this manuscript: [CHAPTER TEXT]. Expand numerals and abbreviations into spoken words. Hold a half-beat after each paragraph. Use a single [pause] before the first line of dialogue only. Keep description slightly slower than speech. Do not add a title sting, theme music, or a second speaker. Return one continuous chapter-open take, dry studio, no crowd bed.
Pro tip: Read the first sentence aloud. If you rush it, the model will too.
Cliffhanger chapter close
2/30✨ What it does
Slows the last two sentences of a chapter and parks a pause before the final line. Keep the file if it ends on breath, not on a spoken chapter card.
Close this chapter in ElevenLabs so the last sentence hangs. Narrator: [VOICE AGE], [GENDER PRESENTATION], dry and precise, original voice only. Pace: [PACE] through the body, then drop 10 percent on the final two sentences. Manuscript: [CHAPTER TEXT]. Spell out times of day. Do not add an extra twist after the last period. Place [pause] after the penultimate sentence, then speak the last line a shade quieter, then stop. No outro music, no 'end of chapter' announcement, no second voice. Dry room. One take that an editor can cut on the breath.
Pro tip: Cut any take that says the words end of chapter. Silence is the button.
Time skip montage read
3/30✨ What it does
Reads dated time-skip beats with a short pause between each expanded date. Keep the take if no whoosh or clock effect sneaks in.
Narrate a compressed time skip in ElevenLabs. Voice age [VOICE AGE], [REGION ACCENT], even timbre. Pace: [PACE] but with a slightly brighter attack so months passing still feel spoken, not sung. Text: [CHAPTER TEXT]. Treat dated headers as spoken dates, fully expanded. Put a short [pause] between each dated beat, never a whoosh or clock SFX. Keep one narrator. Do not invent a montage announcer. Original voice, no famous documentary host. Dry mix so a later music bed can sit under it if the producer wants one.
Pro tip: Write January rather than 1. TTS often trips on numeric dates.
Letter folded into the chapter
4/30✨ What it does
Keeps the book narrator, then lifts formality only for the embedded letter. Keep the file if the accent stays put unless you named one.
In ElevenLabs, read the surrounding narration in the book voice, then shift only for the letter. Book narrator: [VOICE AGE], [PACE], even, original. Letter writer: same mouth but a half-step higher, slightly more formal, still not a new celebrity. Manuscript: [CHAPTER TEXT]. Mark the letter with a [pause] before 'Dear' and after the sign-off. Do not change accent unless the letter writer is named as [LETTER ACCENT]. Expand postal abbreviations. No paper rustle SFX unless the text itself asks the reader to hear paper. One file, two deliveries.
Pro tip: Pause before Dear. A hard cut in the middle of a breath sounds like two files.
Weather interlude between scenes
5/30✨ What it does
Treats a weather bridge as quiet atmosphere with close-mic softness and no extra thunder. Keep the take if the next scene heading is not spoken.
Read this weather bridge in ElevenLabs as atmosphere, not a forecast. Voice: [VOICE AGE], close-mic, [PACE] on the slow side of walking. Text: [CHAPTER TEXT]. Soften plosives. Use [whispers] only if the text already drops to a hush. Name rain, wind, or heat as written; do not add thunder the page does not mention. Original voice, no movie-trailer gravel. Leave headroom so a later rain bed can sit under the take. Stop cleanly when the next scene heading would begin. Dry studio, single narrator.
Pro tip: If the page does not mention thunder, do not ask the voice to boom.
Character Voices
5 promptsVillage elder on the green
6/30✨ What it does
Performs only the elder's lines with open vowels and extra time on place names. Keep the take if no crowd walla appears under the speech.
Cast and read an elder speaking on a village green in ElevenLabs. Speaker age impression: [VOICE AGE], rural [ACCENT], kind, a little worn at the edges, never a famous actor. Pace: [PACE], extra time on place names. Dialogue to perform: [CHAPTER TEXT]. Keep vowels open. Smile on greetings, not on warnings. If a younger listener answers in the text, leave those lines for a later pass; this take is the elder only. No crowd walla. Pronounce local names as [NAME LIST]. Dry, as if a porch, not a hall.
Pro tip: Record the younger answering lines in a separate voice later. Mixing them now muddies the elder.
Court clerk reading the record
7/30✨ What it does
Reads clerk record language at an even bureaucratic pace and only lifts slightly on quoted speech. Keep the file if no gavel or all-rise line was added.
ElevenLabs character: a court clerk, [VOICE AGE], [ACCENT], dry paper timbre, original person not a screen lawyer. Pace: [PACE], even, almost bureaucratic, then a hair slower on the charge. Read: [CHAPTER TEXT]. Spell docket numbers. Do not add gavel hits. Do not impersonate a known judge. If quoted speech from another party appears, keep the clerk's diction and only a slight lift, then return to flat record tone. One take, small room, no reverb cathedral. End on the last recorded line, no 'all rise'.
Pro tip: Spell docket numbers. Digit soup is how TTS loses a jury.
Dock worker at first light
8/30✨ What it does
Gives a salt-edged dock worker read with one optional sigh and holes left for a supervisor. Keep the take if seagulls stay out.
Voice Design plus read for a dock worker at dawn in ElevenLabs. Age feel [VOICE AGE], [PORT ACCENT], low-mid, salt in the throat but not cartoon gravel, not a movie tough. Pace: [PACE] with short sentences. Text: [CHAPTER TEXT]. Keep laugh lines dry. Use [sighs] once if the text shows tired breath, never more. Do not add seagulls. Original voice only. If a supervisor speaks in the same block, skip those lines and leave a one-second hole an editor can fill. Dry warehouse, not a score.
Pro tip: Ask for low-mid, not cartoon gravel. Fake grit turns into a pirate.
Night botanist in the glasshouse
9/30✨ What it does
Walks a greenhouse in a soft awake voice with careful plant names and no insect bed. Keep the take if Latin is only what you listed.
ElevenLabs character voice for a night botanist walking rows. Age [VOICE AGE], [ACCENT], soft but awake, original. Pace: [PACE], curious, never sleepy to the point of slurring Latin. Manuscript: [CHAPTER TEXT]. Pronounce plant names as [LATIN LIST]; if a name is missing, say the common name slowly rather than guessing a famous gardener. Close-mic, slight smile on discoveries, flat on blight. No insect bed. No celebrity nature host. One continuous greenhouse walk, dry, so a later drip can be mixed under.
Pro tip: Paste a pronunciation list. Invented Latin will haunt the whole series.
Traveling merchant in the square
10/30✨ What it does
Runs merchant patter bright, then slows on the story's true cost, skipping buyer replies. Keep the file if no coin effects appear.
Perform a traveling merchant in ElevenLabs. Voice age [VOICE AGE], [TRADE ACCENT], bright without shouting, not a cartoon huckster and not a famous comic. Pace: [PACE] on patter, then slower on the true price of the story, not a shop rate. Lines: [CHAPTER TEXT]. Keep one person. If a buyer replies in the text, skip those replies. Use [excited] only on the first offer, then settle. No market walla, no coin SFX. Original voice. Dry plaza noon, single take a producer can duck under later crowd if needed.
Pro tip: One [excited] tag. Stacking them makes a carnival, not a person.
Dialogue Scenes
5 promptsTwo friends at a kitchen table
11/30✨ What it does
Records only Friend A and leaves silent beats where B answers. Keep the take if no kettle or laugh track sneaks in.
ElevenLabs dialogue pass for two friends. You will generate Friend A only in this take. Friend A: [VOICE AGE], [ACCENT], [PACE], original, not a sitcom star. Kitchen scene text: [CHAPTER TEXT]. Read only the lines tagged A or the first speaker if untagged; skip B. Leave a beat of silence where B would answer. Conversational, contractions intact, no stage laugh. Do not add kettle SFX. Dry domestic room. Later you will run Friend B with a contrasting pitch. This file is A in isolation so the editor can overlap naturally.
Pro tip: Cast B at a different pitch in a second pass. Same mouth twice sounds like an echo.
Argument across a narrow table
12/30✨ What it does
Plays one side of a table fight with clipped consonants and a pause before the last blow. Keep the file if it never turns into a scream.
Read the hotter speaker in a table argument in ElevenLabs. Speaker: [VOICE AGE], [ACCENT], original adult, never a celebrity lawyer. Pace: [PACE] with clipped consonants, not screaming. Scene: [CHAPTER TEXT]. Perform only the lines that belong to [SPEAKER LABEL]. Do not shout through the whole page; land volume on the accusation, then pull back. Use [pause] before the line that ends the fight. Skip the other party's lines. No slap SFX, no crowd. Dry room, close-mic so breath is honest. One take.
Pro tip: Land volume on one accusation. A whole page at full shout is unusable.
Whispered hallway confession
13/30✨ What it does
Whispers only the confession lines and keeps any narration at a low spoken level. Keep the take if no extra footsteps appear.
ElevenLabs close confession. Voice: [VOICE AGE], [ACCENT], original, not a film-star murmur. Pace: [PACE] and actually quiet. Text: [CHAPTER TEXT]. Wrap the confession lines in [whispers] and keep narration (if any) at a low spoken level, not a whisper. Do not add footsteps unless the text names them as words to speak. No second person breathing in the take. Dry hallway, almost no reverb. If a name is sensitive, read exactly the name on the page; do not swap in a famous name. Stop when the confession ends, no swell.
Pro tip: Whisper the confession, speak the setup. Whispering everything turns to mush.
Tavern overlap, one voice only
14/30✨ What it does
Captures a clear storyteller stem with heckles skipped and no generated pub walla. Keep the take if no lute or chart-artist impression appears.
In a noisy tavern scene, record the storyteller only in ElevenLabs. Storyteller: [VOICE AGE], [ACCENT], [PACE], original, not a famous bard. Text: [CHAPTER TEXT]. Skip heckles and bar orders. Do not generate walla; the producer will layer it. Keep the storyteller a little louder and clearer than a real pub, because this is the intelligible stem. No lute. No celebrity impression of a folk singer. Dry. If a song lyric appears, speak it as speech unless [SING LINE] is filled, and even then hum-speak, do not impersonate a chart artist.
Pro tip: Leave walla to the sound bed. Baking it into the voice traps you.
Phone call, near side only
15/30✨ What it does
Records the near side of a call at full body and skips far-side lines for a later filter. Keep the file if no dial tone was added.
ElevenLabs phone scene, near side only. Near speaker: [VOICE AGE], [ACCENT], [PACE], original. Script: [CHAPTER TEXT]. Read the near lines at full body. For far-side lines, skip them entirely so an editor can add a thin filter later, or if you must keep continuity, read them 30 percent quieter and thinner, never as a celebrity voice on the line. No dial tone. No hang-up beep unless the text speaks the word click. Dry room. One adult. Stop on the last near line.
Pro tip: Skip the far side. A fake phone filter inside TTS rarely matches your edit.
Like these prompts? There are full tutorials behind them.
Learn the workflows, not just the prompts. Easy-to-follow tutorials inside AI Academy, and the library keeps growing.
Inner Life
5 promptsGrief walk after the door
16/30✨ What it does
Holds a low even inner walk with at most two sighs and no sniffle effects. Keep the take if it ends at the landmark you named.
Inner narration after a door closes. ElevenLabs voice: [VOICE AGE], [ACCENT], original, not a famous tragic actor. Pace: [PACE], longer vowels, no sniffle SFX. Text: [CHAPTER TEXT]. Keep the body of the walk at a low even level. Use [sighs] at most twice, only where the page shows breath. Do not cry on every clause. Expand times. No second mourner. Dry exterior, as if a quiet street, not a score. End when the walk reaches [LANDMARK], or at the last sentence if that bracket is empty.
Pro tip: Two sighs maximum. Grief that sobs on every clause is unusable in a chapter.
Decision at a rain window
17/30✨ What it does
Gives each listed option a small pause so the fork is audible, with no rain bed baked in. Keep the file if no extra moral was tagged on.
Read an interior decision at a rain window in ElevenLabs. Voice age [VOICE AGE], [ACCENT], thinking speed [PACE], original. Manuscript: [CHAPTER TEXT]. Make questions slightly rising, answers falling. Do not add rain SFX; leave space for a later bed. No celebrity. If the text lists two options, give each option its own small [pause] so the listener can hear the fork. Close-mic, dry glass between the mouth and the storm. One adult. Stop on the chosen line, do not add a moral the page lacks.
Pro tip: Pause on the fork. A smooth blend hides the choice the chapter is about.
Soft dream, then snap wake
18/30✨ What it does
Slows the dream, then snaps to tighter consonants after a pause, with no whoosh. Keep the take if child figures stay in adult-narrator delivery.
ElevenLabs dream-into-wake. Dream voice: [VOICE AGE], slightly slower than [PACE], original, not a film hypnotist. Wake voice: same person, same age, suddenly at [PACE] with tighter consonants. Text: [CHAPTER TEXT]. Use a [pause] at the snap. Do not add whooshes or heartbeats. Do not invent a second dreamer. Keep names as written. Dry. If the dream includes a child character, keep the delivery adult-narrator, never a baby voice. One take that an editor can split on the pause.
Pro tip: Same voice, two speeds. Casting a new mouth for the dream breaks identity.
Memory cut in the present tense
19/30✨ What it does
Cuts to a slightly slower memory after a pause, then returns to present speed on the first present-tense verb. Keep the file if no harp appears.
Perform a present-tense scene that cuts to a memory in ElevenLabs. Present: [VOICE AGE], [ACCENT], [PACE], dry, original. Memory: same mouth, 5 percent slower, a little more air, still not a ghost. Text: [CHAPTER TEXT]. Mark the cut with [pause]. Do not add a harp. Do not impersonate a younger celebrity. If the memory names a place, pronounce it as [PLACE SAYING]. Return to present speed on the first present-tense verb after the memory. One adult. Stop on the last present line.
Pro tip: Five percent slower is enough. Half-speed memories sound drunk.
Counting breaths before sleep
20/30✨ What it does
Speaks a slow pre-sleep count as words with no drone bed, stopping before any nightmare. Keep the take if it stays one person and dry.
Quiet pre-sleep inner count in ElevenLabs. Voice: [VOICE AGE], [ACCENT], [PACE] on the slow edge of speech, original, not a meditation app host. Text: [CHAPTER TEXT]. Speak numbers as words. Do not add a drone bed. Use [whispers] only on the last two counts if the page is already dropping off. No snore. No second sleeper. Dry bedroom, close-mic. If the chapter turns to a nightmare, stop this take at the last calm count so you can start a new file for the scare. One person.
Pro tip: Stop the file at the last calm count. Nightmares need a different take.
Action Reads
5 promptsAlley foot chase
21/30✨ What it does
Shortens phrases on the run and lengthens glances, with no footstep Foley. Keep the take if extra shouted voices stay out unless named.
Narrate a foot chase in ElevenLabs without becoming a sports caller. Voice: [VOICE AGE], [ACCENT], original. Pace: [PACE] with shorter phrases on the run, longer on the glance. Text: [CHAPTER TEXT]. Breathe at commas. Do not add footstep Foley. Do not impersonate a famous action star. Keep one narrator; skip shouted extra voices unless they belong to [RUNNER NAME]. Dry alley. If a turn-by-turn would need a map you do not have, read only what the page says. Stop when the runner reaches [END PLACE] or the last sentence.
Pro tip: Breathe at commas. A chase read that never breathes sounds like a list.
Storm on a small deck
22/30✨ What it does
Punches spoken orders and slows sea description, with no baked-in wind bed. Keep the file if compass points are spoken in full.
Read a small-boat storm in ElevenLabs. Voice age [VOICE AGE], [COAST ACCENT], original sailor mouth, not a movie captain. Pace: [PACE], punchy on orders, slower on what the sea does. Manuscript: [CHAPTER TEXT]. Spell compass points. Do not add wave SFX or wind bed; leave them for the mix. No celebrity. If crew names appear, pronounce them as [CREW LIST]. Dry. One narrator covering both orders and description. Stop after the last order, no 'we survived' tag unless it is on the page.
Pro tip: Leave waves to the sound bed. Voice plus fake wind fights the listener.
Locked room search
23/30✨ What it does
Moves through a room with a pause before each find and no invented hidden door. Keep the take if lock-pick effects stay out.
ElevenLabs search of a locked room. Searcher voice: [VOICE AGE], [ACCENT], [PACE] with caution, original, not a famous detective. Text: [CHAPTER TEXT]. Name objects as written. Do not invent a hidden door. Use [pause] before each find. No lock-pick SFX. Close-mic, dry plaster. If a second person is in the room on the page, skip their lines in this take. Adult only. End on the last object named, not on a solved-case sting.
Pro tip: Pause before finds. Rushing the list makes the room feel empty.
Horse yard arrival
24/30✨ What it does
Keeps dust in the pace and animal sounds out of the throat, with adult narrator even if a child is on the page. Keep the file if no neigh appears.
Narrate an arrival into a horse yard in ElevenLabs. Voice: [VOICE AGE], [ACCENT], [PACE] that can still smell dust, original, not a film cowboy. Text: [CHAPTER TEXT]. Keep animal sounds out of the voice; do not neigh. Pronounce tack terms as [TACK LIST]. No celebrity. Dry yard, noon. If a child appears in the text, keep the narrator adult and do not switch to a high cartoon voice. One take. Stop when the rider dismounts or at the last sentence. Leave hoof Foley to the editor.
Pro tip: Do not neigh. TTS animal imitations wreck a yard scene.
Kitchen scramble, dropped tray
25/30✨ What it does
Lands the drop, then works in short clauses with no crash effect or laugh track. Keep the take if extra staff stay silent unless named.
Read a kitchen scramble after a dropped tray in ElevenLabs. Voice: [VOICE AGE], [ACCENT], [PACE] with quick short clauses, original cook, not a celebrity chef. Text: [CHAPTER TEXT]. Do not add crash SFX. Do not shout the whole page; land the drop, then work. Skip extra staff lines unless labeled [COOK NAME]. Dry tile. Expand temperatures into spoken numbers plus degrees. One adult. Stop when the tray is righted or the page ends. No laugh track.
Pro tip: Speak temperatures as words. 400 reads as four hundred, not four zero zero.
Go from copy-pasting to actually mastering AI.
AI Academy: hands-on tutorials on ChatGPT, Claude, Midjourney, and many other tools. New tutorials are added regularly.
Nonfiction Tone
5 promptsMemoir at the kitchen sink
26/30✨ What it does
Keeps a first-person memoir in an original author-voice with years expanded and no music sting. Keep the take if names stay as written.
Nonfiction memoir read in ElevenLabs. Narrator is the author-voice: [VOICE AGE], [ACCENT], [PACE], original, not a famous memoirist. Text: [CHAPTER TEXT]. Keep first person. Do not sweeten it into a novel. Expand years. If a relative is named, read the name as written; do not swap a celebrity. Dry domestic. Slight smile on ordinary details, flat on hard facts. No music sting. One take the author can approve line by line. Stop at the section break.
Pro tip: Approve names before you generate. A swapped relative is a rights mess.
History footnote, not a lecture
27/30✨ What it does
Speaks years in full in a friendly footnote tone with no battle effects. Keep the file if missing figures were skipped, not filled.
Read a history aside in ElevenLabs as a footnote a friend might say, not a podium. Voice: [VOICE AGE], [ACCENT], [PACE], original, not a famous presenter. Text: [CHAPTER TEXT]. Speak years in full. If a figure is missing, skip it rather than filling a famous statistic. No battle SFX. Dry. Pronounce names as [NAME SAYINGS]. One narrator. Do not add a class-dismissed closer. Stop on the last sourced sentence.
Pro tip: Skip a missing year. A guessed date will date the whole book.
Field note from a trail
28/30✨ What it does
Walks a field note at walking pace with species names locked and no added bird calls. Keep the take if units are spoken.
Nature field note in ElevenLabs. Voice: [VOICE AGE], [ACCENT], walking [PACE], original, not a television ranger. Text: [CHAPTER TEXT]. Keep species names as [SPECIES LIST]. Do not add bird calls. Do not impersonate a famous naturalist. Dry trail, close-mic, breath allowed. If a measurement is in the text, speak the unit. One adult. Stop at the last observation. Leave wind for the mix.
Pro tip: Paste species pronunciations. Guessed Latin loses trust on a trail.
How-to hands, one step at a time
29/30✨ What it does
Speaks numbered steps as words with a pause after each and no tool effects. Keep the file if missing measurements were skipped.
Instructional nonfiction in ElevenLabs. Teacher voice: [VOICE AGE], [ACCENT], [PACE] a learner can follow, original, not a celebrity maker. Text: [CHAPTER TEXT]. Number steps as words: step one, step two. Pause after each step. Do not add tool SFX. If a measurement is missing, skip the step rather than inventing a famous shortcut. Dry bench. One adult. Stop after the last listed step, no subscribe closer.
Pro tip: Pause after each step. Learners rewind; a rush forces them to.
Archive interview, editor stem
30/30✨ What it does
Records only interviewer framing with guest answers skipped for real tape. Keep the take if no station ID or theme appears.
Read the interviewer stem of an archive clip in ElevenLabs. Interviewer: [VOICE AGE], [ACCENT], [PACE], original journalist mouth, not a famous anchor. Questions and framing: [CHAPTER TEXT]. Skip the guest answers so the real guest tape can sit in. No theme music. No station ID. Dry. If a date is in the intro, speak it in full. One adult. Stop before the guest would begin. This is a bed for an editor, not a whole show.
Pro tip: Stop before the guest. Over-reading the answer traps the archive.
Free tool
Prompt Optimizer
Turn a rough idea into a structured, professional AI prompt.
Frequently Asked Questions
Prompts are the starting line. Tutorials are the finish.
A growing library of hands-on tutorials on ChatGPT, Claude, Midjourney, and many other AI tools. New tutorials are added regularly.