Synthesia Flux 3 B-roll: 20 seconds with audio, and the Sume route

Synthesia's updates page says that on September 16, 2026 it added B-roll generation with Flux 3 from Black Forest Labs, creating up to 20 seconds of B-roll with natively synced audio inside the video editor. Sume does not generate B-roll inside an avatar job: you make the talking avatar clip and the B-roll clips as separate jobs and combine them with Timeline 1.0.

What Synthesia says

The entry is on the Synthesia product updates page, read 2026-10-02. A neighbouring September 25 entry adds 720p, 1080p or 4K choices for generated B-roll.

• The page describes this as an editor feature, so the clip lands in your Synthesia scene directly.

How the same result is built on Sume

Sume's avatar video job produces the talking clip: 4 to 60 seconds, 720p, one avatar and one shared scene. B-roll comes from the video generation models, which are separate jobs. The Models page notes that wordless beats, B-roll and product motion are not Fabric or avatar clips.

You then join the pieces. Timeline 1.0 sequences ordered video slots over one audio spine, and Timeline compose puts one still and one video on screen together, up to 300 seconds.

A practical split

Keep the voice track continuous. Render the avatar clip first so the speech timing is fixed, then cut away to B-roll over the spoken lines, and keep the original audio under the cutaway.

Check the audio on any B-roll clip you reuse. Synthesia's page says its Flux 3 clips arrive with synced audio; on Sume, decide per clip whether to keep that sound or let the avatar's narration run underneath.

• Synthesia: one editor, B-roll generated in the scene.
• Sume: separate jobs, joined by your timeline document.
• Neither page here lists a price for the B-roll clip, so compare live rates before choosing.