How to make an explainer video with AI: two ways

To make an explainer video, write a short script, split it into beats (for example the problem, how it works, and what to do next), give each beat a picture, record the voiceover, and edit the beats together with captions and music. With AI on Sume you pick one of two routes: an avatar presenter speaks the beats in one multi-scene Avatar 1.0 video of 4–60 seconds, or a text-to-speech voiceover runs under generated B-roll in a Timeline 1.0 render. Longer explainers are several parts joined on one timeline.

Facts come from Sume's Generate avatar video, Models overview, Timeline 1.0, and Video captions docs and the Sume API reference, read on 2026-09-27; anything called current behavior is read from Sume's code. Each mechanism has its own post, linked below.

Can you make an explainer video with AI?

Yes, for the parts that are spoken or illustrated: the voice, a presenter, B-roll, music, and burned-in captions. What generation doesn't give you is footage of your real product: a generated clip illustrates, it doesn't record. Screen walkthroughs come from your own recorder and editor, because Timeline takes only Sume-hosted files and Sume's public API has no route for uploading a local file.

Should the explainer use a presenter or a voiceover?

Choose by what the viewer should see while listening. A presenter suits a talk-to-camera explainer; B-roll suits one that shows many different things.

How do I make the presenter version?

Send the beats as ordered video_inputs scenes instead of one script. A spoken beat is voice.type: "text" with its script; a pause for a visual is voice.type: "silence" with a duration. Current execution supports one avatar per video and expects scene backgrounds to resolve to one shared scene, so the presenter can't stand in front of a different screenshot for each beat. To show B-roll during a beat, cut away on Timeline while the avatar's voice keeps playing, as in Add B-roll to a talking-head avatar video. Scene fields are in Multi-scene avatar video API.

curl -X POST https://api.sume.com/v1/avatar-1.0/talking-video \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: explainer-presenter-001" \
  -d '{
    "avatar_handle": "acme",
    "aspect_ratio": "16:9",
    "video_inputs": [
      { "id": "problem", "voice": { "type": "text", "script": "Invoices get lost when nobody owns them." } },
      { "id": "visual", "voice": { "type": "silence", "duration": 3 } },
      { "id": "how", "voice": { "type": "text", "script": "Acme gives every invoice one owner and a reminder." } }
    ]
  }'

How do I make the voiceover version?

Voice the script with TTS 1.0, generate one clip per beat with POST /v1/videos, and lay the clips over the narration in one Timeline 1.0 render with a music bed. Keep talking people out of the B-roll: Sume's docs say video models do not lip-sync to generated TTS or to a later voice-over. Faceless video API walks through every call, including captions from the TTS word timings.

How do I make an explainer longer than 60 seconds?

Split the script into parts of up to 60 seconds and make each part. Caption each part while it is short: today a caption job refuses a source longer than 60 seconds. Then join the parts in one Timeline 1.0 render. Today a render's sound is only its audio spine and soundtrack, so extract each avatar part's audio with POST /v1/audio-detach and pass the files in order as audio.parts[] (up to 20), with the matching videos in video[]. How to make an AI avatar video longer than 60 seconds shows the join.

What are the limits, and what does it cost?

• Avatar 1.0 speaks English only in current code. For another language, voice the script with TTS 1.0 and language, then use VEED Fabric 1.0 to lip-sync a presenter still to that audio (lip sync API).
• Timeline and the media tools take only your workspace's media.sume.com files, such as earlier Sume outputs.
• Avatar video bills per second by quality tier: $0.184/s standard, $0.245/s plus, $0.55/s max (no product image). Narration is $0.0475 per 1,000 characters, and each render is listed at $0.10 per output minute, each plus a 5.5% agent fee by default.