Voiceover Builder

ElevenLabs v4 voiceovers with direction and audio tags · 32 fields · 5 outputs

Library

Filled only

Project Type

Your Script

Audience

Goal

Length

Voice Gender

Voice Age

Accent

Voice Timbre

Personality

Reference Feel

Base Emotion

Primary Audio Tag

Secondary Audio Tag

Emphasis Style

Pause Strategy

Emotional Arc

Pace

Words Per Minute

Script Handling

Pronunciation Notes

Language

ElevenLabs Model

Voice Source

Speakers

Stability Setting

Output Format

Post Production

Sync Target

Platform

Take Plan

Rights Check

🔒 Preview mode
Try every field. Unlock Voiceover with Pro to copy and save the prompts.

✨ Enhance with AI
Rewrite the main prompt into 3 polished variations. Pro feature.

01Main Voiceover Brief

VOICEOVER BRIEF: Explainer Video for a General Consumer audience GOAL: Inform and Explain LENGTH: 30 to 60 Seconds at 150 WPM Conversational VOICE: Female Adult 30s with a Neutral American accent TIMBRE: Warm and Rich PERSONALITY: Warm Friendly Guide in the style of a Friendly YouTuber EMOTION: Warm and Welcoming with a Steady Throughout arc PACE: Measured Conversational EMPHASIS: Emphasise Key Nouns PAUSES: Natural Sentence Pauses LANGUAGE: English MODEL: Eleven v4 Most Emotive using Voice Design from Description SPEAKERS: Single Narrator at Natural Balanced stability DELIVER: MP3 High Quality for YouTube

02Voice Design Prompt

VOICE DESIGN PROMPT - PASTE INTO THE VOICE DESCRIPTION BOX A Adult 30s Female voice with a Neutral American accent, Warm and Rich in tone, sounding Warm Friendly Guide and delivering in the manner of a Friendly YouTuber - speaking at a Measured Conversational pace with a Warm and Welcoming quality. DESCRIPTION BOX RULES Keep the description between about 20 and 1000 characters Describe the voice itself, not the script it will read Age, accent and timbre matter most - name all three Generate several previews and save the one you want to reuse Use a fixed seed if you need the same voice again later VOICE SOURCE: Voice Design from Description NOTE: Professional Voice Clones are not fully optimised for v3 - test before committing NOTE: ElevenLabs Default voices expire at the end of 2026 - save a designed voice you own

03Tagged Script

TAGGED SCRIPT - PASTE INTO THE TEXT BOX [warm] TAG PLACEMENT GUIDE Open with [ warm ] to set the base delivery Use [ softly ] before any line that needs a shift Place a tag immediately before the words it should affect, not after Tags are lowercase and inline - they are never spoken aloud Two or three tags in a passage works. A tag on every line does not. Tag response varies by voice - test yours before a long read SCRIPT HANDLING: Use My Script As Is CHUNK LONG SCRIPTS - v3 caps at about 5000 characters per request

04Delivery and Pacing Brief

BASE EMOTION: Warm and Welcoming ARC: Steady Throughout across the whole read PACE: Measured Conversational targeting 150 WPM Conversational EMPHASIS: Emphasise Key Nouns PAUSES: Natural Sentence Pauses PERSONALITY: hold Warm Friendly Guide from the first word to the last AUDIENCE: pitch the read for a General Consumer listener GOAL: the read must Inform and Explain PUNCTUATION IS DIRECTION - commas, full stops and ellipses change the timing WRITE FOR THE EAR - short sentences read better than long clauses READ IT ALOUD FIRST - if you stumble, the model will too STABILITY: Natural Balanced - lower is more expressive, higher is more consistent

05Production and Rights

PRODUCTION AND RIGHTS PLAN MODEL CHOICE: Eleven v4 Most Emotive V3 for expressive reads with tags. Multilingual v2 for steady neutral narration. Flash for real time. OUTPUT: MP3 High Quality TAKE PLAN: Re roll Problem Lines Only FIX LINES NOT WHOLE READS - regenerate only the problem sentence and splice it POST: Keep Breaths Natural SYNC: Sync to Video Cuts MIX NOTE: if music sits under this, duck it around the 1 to 4 kHz range PLATFORM: YouTube RIGHTS: Check Your Plan Allows Commercial Use NEVER CLONE A VOICE WITHOUT THE OWNER EXPLICIT PERMISSION - this is the line that matters most Commercial use depends on your plan tier - verify before selling work made with it Some platforms and regions require disclosure of synthetic voice - check yours Keep records of your prompts, voices and edits

📘 Pro Guide

Model Guide

Eleven v3Most expressive. Audio tags, multi speaker, 70 plus languagesDefault for content
Multilingual v2Still preferred by many for steady neutral narrationAudiobooks and courses
Flash v2.5Very low latency for real time and agent use, fewer languagesLive applications
v3 is not real timeIt caps around 5000 characters per request - chunk long scriptsPlan for it
Clone qualityProfessional Voice Clones are not fully optimised for v3 yetTest first
Default voicesElevenLabs Default voices expire at the end of 2026Save your own

Audio Tags

Lowercase in bracketsTags go inline in the script and are never spoken aloudFormat
Place before the wordsA tag affects what follows it, not what precedes itPlacement
Do not over tagTwo or three across a passage works. One per line does not.Restraint
Voice dependentTags respond differently across voices - test yoursKnown variance
Emotion tagsexcited, sad, angry, curious, nervous, confident, warmFeeling
Performance tagswhispers, laughs, sighs, exhales, chuckles, clears throatDelivery

Voice Design

Describe the voiceNot the script - describe age, accent, timbre and mannerCore rule
Name three thingsAge, accent and timbre carry the most weightWhat matters
Length windowRoughly 20 to 1000 characters for the descriptionConstraint
Generate previewsMake several, listen, then save the one you wantWorkflow
Save and reuseA saved voice is available across all their toolsConsistency
Use a seedIf you need the same voice again, keep the seedRepeatability

Writing for the Ear

Read it aloudIf you stumble reading it, the model will stumble tooBest test
Short sentencesLong clauses lose the listener and confuse the phrasingClarity
Punctuation is directionCommas, full stops and ellipses change the timingControl
Spell out numbersWrite dates, prices and acronyms the way you want them saidAccuracy
Watch brand namesUnusual spellings get mispronounced - test them alone firstPronunciation
Fix lines not readsRegenerate the problem sentence and splice it inEfficiency

Rights and Ethics

Never clone without consentCloning a real voice without explicit permission is the lineNon negotiable
Check your planCommercial use rights depend on your subscription tierBefore selling
Disclosure rulesSome platforms and regions require synthetic voice disclosureEvolving
Keep recordsSave your prompts, voice IDs and editsProtection
Voice actorsFor high stakes brand work, a human is still often the right callHonest advice

Preset ideas

Explainer NarratorFemale Adult 30s
Documentary VoiceMale Middle Aged
Punchy Ad ReadFemale Young Adult
Audiobook NarratorMale Adult 40s
Meditation GuideFemale Adult 40s
UGC Creator ReadFemale Young Adult
Corporate TrustMale Adult 40s
Kids Story VoiceFemale Adult 30s
Trailer VoiceMale Middle Aged

Model status (checked Nov 2026)

Eleven v4Launched 28 September 2026: ElevenLabs' most emotive model, 90+ languages, audio tags and direction prompts. v4 Turbo (about 100 ms) is built for real-time voice agents.Voice clones made before v4 should be retrained for best results.
Older modelsEleven v3 and Multilingual v2 remain available. Flash v2.5 is still the lowest-latency option. Turbo v2.5 is deprecated.