Avatar Video Builder

Talking-head scripts and settings for HeyGen, Synthesia and D-ID · 32 fields · 5 outputs
Library
30 / 32 FIELDS ACTIVE

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

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

01Main Avatar Prompt
AVATAR VIDEO: Product Explainer for a General Consumer audience PRESENTER: 30s Female in Business Casual EXPRESSION: Friendly Approachable with Subtle Natural Gestures VOICE: Warm Female Alto with a Neutral American accent in English DELIVERY: Natural Conversational FRAMING: Chest Up at Straight On Eye Level BACKGROUND: Blurred Office lit with Soft Even Studio ON SCREEN: Captions TOOL: HeyGen in Text to Avatar Video for 30 to 60 Seconds OUTPUT: 9:16 Vertical at 1080p Full HD AVOID: Uncanny Stiffness
02Avatar Appearance Brief
03Voice and Lip Sync Brief
04Framing and Set Brief
05Production and Consent
📘 Pro Guide
Avoiding the Uncanny Valley
  • Micro movement beats resolutionA perfectly still face reads as dead, however sharp it isNumber one fix
  • Irregular blinkingMetronomic blinking is the fastest tell that it is syntheticDetail
  • Gestures on the beatHands moving against the words is worse than no hands at allTiming
  • Match wardrobe to settingA suit in a bedroom breaks the illusion instantlyCoherence
  • Blurred backgrounds helpDepth hides more artifacts than a flat colour doesPractical
  • Shorter is saferArtifacts accumulate - long unbroken takes drift moreLength
Lip Sync
  • Watch the final secondsDrift almost always appears at the end of a long takeWhere it fails
  • Break at breath pointsShort sentences sync far better than long clausesScripting
  • Test the hard wordsNumbers, acronyms and brand names mispronounce most oftenAccuracy
  • Fix the sync, not the videoIf the voice is right and the mouth is wrong, redo the syncEfficiency
  • Match face to voiceA young presenter with a much older voice reads as wrongCasting
  • Upload audio for controlSupplying your own audio gives tighter control than text inputTechnique
Framing That Builds Trust
  • Eye line into the lensLooking slightly past the camera reads as evasiveTrust
  • Controlled headroomA floating head in empty space looks amateurComposition
  • Vertical cropsKeep the face upper middle, clear of the caption zoneSocial
  • Clear the mouthLogos and lower thirds must not sit over the mouth or handsOverlays
  • Captions alwaysMost viewers watch on mute, whatever the platformUniversal
  • Consistent series lookSame avatar, wardrobe and background across a setBranding
The Real Efficiency
  • Update script, keep avatarThis is the whole point of the formatCore benefit
  • Batch localisationOne avatar, many languages, no reshootScale
  • Template itBuild once, then swap only the wordsWorkflow
  • Series consistencyViewers accept a synthetic presenter more once familiarAudience
  • Know the limitsFor high stakes brand work a real person still usually winsHonest
Consent and Disclosure
  • Written permission alwaysNever create an avatar of a real person without itNon negotiable
  • That includes employeesStaff must consent explicitly, not by assumptionCommon mistake
  • Never public figuresUsing a known face without permission is a serious problemLegal risk
  • Disclosure is tighteningPlatform and regional rules are changing - check yoursEvolving
  • Extra care on adviceMedical, financial and legal content carries added dutyResponsibility
  • Keep your recordsSave consent, licences and generation historyProtection
View prompts ↓