How HeyGen Prompts Work
HeyGen lets you control three things: how AI avatars look, how they move, and what they say. Effective prompts treat each as a separate concern. Use one prompt for the avatar appearance, one for the script, and configure voice and pacing separately. Mixing all three into one prompt produces inconsistent results.
For avatar looks, structure prompts into image type, main subject, background scene, and composition style. For scripts, use the Role + Audience + Goal + Duration + Tone formula. For motion, specify energy level and gesturing style directly in HeyGen's settings panel.
Prompt Formula for Avatar Looks
HeyGen's avatar look guide uses: Image type + Main subject + Background scene + Composition & aesthetics
"Studio headshot photo of a professional woman in her 30s wearing a blazer, neutral gray background, mid-range composition, natural lighting, confident but friendly expression, highly detailed, realistic."
Key variables to adjust:
- Setting: studio, office, café, classroom, outdoor, abstract background
- Lighting: natural, studio, soft, dramatic, warm, cool
- Framing: close-up, mid-range, full body, over-the-shoulder
- Expression: confident, warm, authoritative, approachable, serious
25+ Avatar Look Prompts
- Professional woman, 30s, blazer, neutral background, mid-range, realistic
- Middle-aged man with glasses, modern office, soft lighting, blurred background
- Young Black woman, casual streetwear, colorful city wall, warm tones
- South Asian man, button-down shirt, light beige background, professional look
- Woman, 40s, curly hair, bright blouse, home office with plants
- Young man, hoodie, headphones, dark background, tech creator vibe
- Woman in café with laptop, warm light, shallow depth of field, relaxed
- Older man, gray hair, kind eyes, soft blue background, close-up, gentle smile
- Woman, business attire, corporate office, daylight from windows, confident
- Fitness coach, sportswear, gym setting, strong lighting, energetic expression
- Teacher, cardigan, warm classroom background, friendly and approachable
- Doctor, white coat, clinical background, calm and trustworthy expression
- Tech CEO, smart casual, modern office with city view, authoritative yet relaxed
- Customer service rep, headset, branded polo, neutral background, warm smile
- Financial advisor, suit and tie, marble desk background, sophisticated
Script Prompt Formula + 30+ Ideas
Use this template with ChatGPT or another LLM to generate your HeyGen script, then paste it in:
"Role: [who the avatar represents — e.g. customer success manager]. Audience: [who is watching — e.g. new enterprise customers]. Goal: [what they should know/do after watching]. Length: [duration — e.g. 90 seconds]. Structure: Hook (10s) → Main content (60s) → CTA (20s). Tone: [e.g. professional, warm, energetic]. Keep sentences under 15 words. Write for speaking, not reading."
HeyGen Best Practices 2026
Separate avatar, script, and voice prompts
Three distinct prompts produce far better outputs than one blended request. Define the look first, finalize the script second, and choose the voice last.
Write scripts at 130-150 words per minute
Natural spoken delivery is slower than you think. A 60-second video needs roughly 130-150 words — not 200+. Overcrowding the script makes the avatar sound rushed.
Use US English for highest-quality voice output
HeyGen's voice quality is highest in US English. If you need other languages, generate in English first, verify quality, then switch to the target language using HeyGen's translation feature.
Add punctuation for natural pacing
Commas create natural pauses. Periods signal a breath. Short sentences under 15 words sound confident and clear. Avoid complex clauses that force the avatar into unnatural phrasing.
Test a 20-second segment before generating the full video
HeyGen charges credits per generation. Always test a short segment with your chosen avatar and voice before committing to a full 3-5 minute video.
Frequently Asked Questions
What is the best prompt structure for HeyGen avatar looks?
Image type + Main subject + Background scene + Composition and aesthetics. Be specific about lighting, framing, expression, and demographic details. More specificity equals more consistent results across multiple generations.
How many languages does HeyGen support?
HeyGen supports 140+ languages with automatic translation and lip-sync matching. Ideal for global training programs and multilingual marketing content without re-recording.
Can I use my own face as a HeyGen avatar?
Yes. HeyGen offers Instant Avatar (upload a short video of yourself) and Studio Avatar (professional quality with a dedicated recording session). Both allow you to create a personal AI clone for unlimited video generation.
What video length works best for HeyGen?
Under 3 minutes for social content, 2-7 minutes for training modules, 90 seconds or less for ads and pitches. Keep the avatar moving with gestures and scene cuts to maintain engagement past the 2-minute mark.