How to split a training lesson into avatar clips under 60 seconds

Split the lesson at its natural breaks into parts of 30 to 55 seconds, render each as its own avatar video, then join them. Sume accepts a script only if its estimated duration is 4-60 seconds, so a six-minute lesson is about seven jobs, not one.

The HeyGen guide on interactive enterprise video, read 2026-10-06, notes that training platforms add quizzes and branching on top of video. Splitting also helps there, because each clip can sit before one question.

How long each part can be

Planning on 84 to 154 words per part keeps every clip safely inside the window, with room for a pause.

A splitting rule

Write the lesson as numbered ideas, one idea per clip. Count words per idea and cut any that pass about 150. Keep the same avatar handle and the same scene prompt in every clip so the presenter looks the same throughout.

Then join clips on Timeline 1.0, which is priced at $0.10 per ceil of output minute, so a six-minute assembly is about $0.60.

If a lesson has an exercise, end the preceding clip on the question, and let your learning tool show the exercise. Do not ask the avatar to wait; a rendered clip has no way to react.

When the lesson changes, only the affected clips need a re-render. That is the main practical reason to split: one corrected sentence costs one clip, not the whole course.

Checklist

Use this as a pre-flight list before you spend anything.

• One idea per clip, 84-154 words.
• Same avatar handle and same scene prompt for all parts.
• Preview the first frame once; reuse the look for the rest.
• Captions on, because lessons are often watched muted.
• Name clips in order, lesson-01 onward, so the timeline is easy to assemble.

Cost of the whole lesson

Seven parts of about 50 seconds each is about 350 seconds of video, which is roughly $64 at standard, $86 at plus or $193 at max, before the Timeline step. Judge the tier on one part first; a lesson reused for a year can justify plus.

A cheaper route for the first draft is standard tier on every part. Promote only the clips a learner will watch most to plus or max.