Add titles, lower-thirds and callouts (motion graphics)

Goal: add on-screen graphics to an edited video — a title at the start, a "lower-third" name caption when someone is introduced, callouts for key points — placed where they do not cover faces or busy parts of the picture, at moments that match what is being said.

ReelBolt's motion-graphics AI (MotionGraphicsPlanner) designs each graphic as a small Remotion animation that it writes and renders itself, with a transparent background, and the VideoCompile step lays it over the footage. Plain text on a semi-transparent box is the fallback when an animation is not appropriate or fails to render.

Prerequisites for motion graphics

• The video uploaded to the project.
• A chat AI provider, and the Remotion sandbox service running (standard installation).
• Recommended: a transcription provider, so graphics line up with what is said (for example a name appears when the person is introduced).

The motion graphics workflow

Add the Footage edit with titles and callouts template (video-derush-edit-graphics) and select your video in step 1.

#	Label	Step type	Agent
1	Analyze source video	VideoAnalyze	VideoTransform (not an AI)
2	Decide which spans to keep	Agent	VideoStoryEditor
3	Plan motion graphics overlays	Agent (context FullWorkflow)	MotionGraphicsPlanner
4	Compile edited video with graphics	VideoCompile	VideoTransform (not an AI)
5	Review edit quality	ReviewLoop	VideoReviewAgent

Step 1 finds good spots for graphics:

{"version": 1, "source": {"kind": "ProjectFile", "projectFileId": "PUT-THE-FILE-ID-HERE"},
 "emitOverlayPlacements": true}

Each spot is offered to the AI as an id (p0, p1, ...) with its area of the screen (LowerThird, UpperThird or CenterBand), a suitability score, a hint for light or dark text, and the moment it belongs to. Long, still shots get several spots at different moments (maxTimeSlicesPerRegion, default 3). Limits: maxPlacementsPerShot 2, maxPlacements 40.

Step 3 must see both step 1 (the spots) and step 2 (what was kept), so its context is FullWorkflow. Each spot is marked with whether its moment survived the edit, and the AI prefers those.

Step 4:

{"version": 1, "decision": {"from": "Step", "stepOrder": 2}, "analysisStepOrder": 1,
 "enableGraphics": true, "graphicsPlan": {"from": "Step", "stepOrder": 3},
 "transitionPolicy": "Auto", "programFadeInMs": 500, "programFadeOutMs": 800,
 "programAudioFadeInMs": 300, "programAudioFadeOutMs": 900, "minSegmentMs": 800}

graphicsPlan must point at the planner step by number. "Previous" would also work here, but the numbered form keeps working if you add steps.

How to run and steer the graphics

When you click Run, describe what you want in Your brief:

Add a title card "Spring Launch 2026" at the start, a lower-third "Maria Lopez, Head of Product" when Maria introduces herself, and at most two callouts for the key numbers. Bold, modern style, white text with a purple glow.

The AI chooses only among offered spots and picks a duration word for each graphic: Short (1.5 s), Medium (3 s) or Hold (6 s) by default. It never chooses times or positions itself; ReelBolt places the graphic exactly from the chosen spot.

Step 3 is the slowest step: for each graphic the AI writes a Remotion animation, checks it builds, and renders it.

What to check after adding graphics

• Step 1 result: meta.offeredPlacementIdCount above 0, and a view.placements list.
• Step 3 result: the overlays list. A graphic with a renderedAssetStorageKey is a rendered animation; one with only text/subtext is a plain text box (its reason should say why).
• Step 4 result: the graphics section reports how many graphics were applied, and lists any that were dropped with a reason (for example a spot whose moment was cut out of the edit).

Settings for plain-text graphics

These VideoCompile settings only affect the plain-text fallback boxes (defaults in brackets): overlayFontSizePct (5), overlayBoxHeightPct (16), overlayBoxWidthPct (82), overlayFontColor ("white"), overlayBoxColor ("[email protected]"), maxOverlayTextChars (80), maxOverlaySubtextChars (60). Durations: overlayShortMs (1500), overlayMediumMs (3000), overlayHoldMs (6000), with a fade of overlayFadeMs (300). maxOverlays (20) caps how many graphics are applied.

Variations of motion graphics

• Panel of artists: the Footage edit with titles, designed by a graphics team template (video-derush-edit-graphics-room) uses a GraphicsRoom step instead of step 3. A layout artist, a timing artist, a copy artist and an insert artist discuss the options, then a MotionGraphicsDirector writes the plan and renders the graphics. Its view must point at the analysis step: {"version": 1, "view": {"from": "Step", "stepOrder": 1}}.
• Graphics inside a green screen: the same AI can also fill a green phone or laptop screen with a rendered scene that follows the screen. See Tracked screen insert recipe.
• No cutting, just graphics: tell the editor in the request to keep the whole clip.

Pitfalls with motion graphics

• No graphics at all: step 1 was missing "emitOverlayPlacements": true, step 4 was missing "enableGraphics": true, or graphicsPlan pointed at the wrong step. Also, choosing zero graphics is allowed: the AI is told to prefer none over clutter, so ask for them explicitly.
• Only plain text boxes: rendering failed or the sandbox was not available. The overlay's reason usually says which.
• A graphic appears at the wrong moment for the words: without transcription the AI cannot know what is said when. Add a transcription provider.
• GRAPHICS_REQUIRE_REENCODE: graphics need "mode": "Reencode" (the default).
• Long step 3: expected. Each rendered graphic takes several tool rounds; plan for minutes per graphic.