Avatar Presenter Training Video Over 60 Seconds: Split Into Jobs

The short answer

Sume accepts avatar video scripts that it estimates at 4 to 60 seconds, so a 3-minute presenter video is three jobs of up to 60 seconds, joined in Timeline 1.0. At standard quality ($0.184/s) 180 seconds is $33.12, at plus ($0.245/s) $44.10 and at max ($0.55/s) $99.00, plus one timeline render of 3 minutes ($0.30) and a one-time $0.95 to create the avatar.

How to split

Split the script at topic breaks, not at a word count, so each job ends on a complete thought. Keep an eye on the estimate: docs say Sume accepts 4 to 60 seconds inclusive, and a script that estimates longer must be made shorter or split. Give the three jobs the same avatar_handle, aspect_ratio (default 9:16) and quality, so the person looks the same across the cuts. Resolution is 720p.

For a multi-scene job, use ordered video_inputs. A voice.type: silence beat with a duration is a pause with no speech; use one between two points that need breathing room. The current execution supports one resolved avatar for each final video and expects the scene backgrounds to resolve to one shared scene.

Review before the full run

Use an avatar video preview to see first-frame stills before you pay for a render, and generate the video from the preview id once you approve the framing. Stills only show the first frame, so they do not prove lip movement or pacing; watch the first job in full before you commit the other two.

Inline captions on the avatar job are an add-on to the estimate and are burned after generation; a failed caption stage is a soft failure and leaves a clean video_url. For three jobs, caption once on the stitched file instead if the total is under 60 seconds of speech per caption job.

One more budget line: a failed or rejected take. At plus quality, a full 60-second retry is $14.70. If you plan for one retry per three jobs, add that to the 180-second line, and consider standard quality for the first read-through, since 60 seconds at standard is $11.04.

• Use only an avatar from a person who consented to it.
• Disclose that the presenter is synthetic where your company policy or local rules ask for it.
• Pick plus for the first job, then decide whether the other two need max.