How to create your first Synthesia video with an AI avatar

Open Google Docs, Notion, or any text tool.

Write 150-200 words for a 60-90 second video. Synthesia avatars read at ~150 words/min.

Read it out loud. Cut any sentence over 20 words. Cut any paragraph over 3 sentences.

Write for the ear, not the eye. Conversational rhythm beats written-prose rhythm every time.

Mark intended pauses with [pause] — Synthesia respects these as natural silence.

Synthesia ships 200+ stock avatars across ages, ethnicities, styles.

For sales outreach: clean, warm, age-matched to your target buyer.

For training: more authoritative, professional dress, neutral background.

For social media: more casual, animated, lifestyle background.

Filter avatars by tone (corporate, casual, friendly) and shot (close-up vs medium).

Preview 5-7 avatars with a 1-line snippet of your script before committing.

For each avatar, Synthesia suggests 2-4 voice matches. Start with those.

Adjust pace: 90% default speed reads more natural for most marketing video.

Adjust emphasis: highlight 3-5 words per paragraph for emphasis using the editor controls.

Add SSML if needed: <break time="500ms"/> for pauses, <emphasis> for stress. Synthesia editor exposes these visually.

Test the first 10 seconds before committing. Voice fit becomes obvious or wrong fast.

Scene 1 (0-10s): avatar intro with strong hook. Avatar on screen.

Scene 2 (10-25s): cut to screen recording, product shot, or supporting visual. Avatar may stay as a small picture-in-picture.

Scene 3 (25-50s): back to avatar, main content. Add text overlays for key points.

Scene 4 (50-75s): supporting visual or chart. Show what you are saying.

Scene 5 (75-90s): avatar close-up for CTA. Direct eye contact, slower pacing.

Synthesia ships scene templates — use them as starting points, not finished structure.

Synthesia auto-generates subtitles. Click "Add subtitles" → review.

Review every line for misspelled product names, acronyms, brand terms. Auto-generation gets these wrong often.

Style: white text with black background or shadow. Readable on bright phone screens.

Preview on mobile dimensions (9:16 or 16:9 zoomed) before final render.

Reading speed: subtitles should match avatar pace. Reset timing if anything reads too fast or too slow.

Hit Render. 1080p typically takes 5-15 minutes depending on length.

Watch the rendered output on the device your audience will use (phone for social, desktop for B2B).

Note timing of any "uncanny" moments — usually mouth shape on specific words. Adjust pacing or rephrase.

Check audio levels, music balance, intro/outro transitions.

Re-render. Most production-quality videos go through 2-3 render iterations before publication.