Fix the Giant Text Box on Upgraded Captions in Premiere Pro
By the Caption Plug team · Published June 13, 2026 · 5 min read
Captions upgraded to graphics in Premiere Pro come in as paragraph text: each clip inherits the caption track's full-width text container, so you get a giant invisible bounding box with the anchor point at its center - not the word's. Scale and rotate animations then pivot around the wrong spot and the text warps, slides or orbits instead of popping in place. The fix is converting each layer to point text and re-centering the anchor. Here's exactly how, plus the realistic time cost when there are two hundred of them.
What's actually wrong with the upgraded layer
Premiere has two kinds of text layer. Point text (what the Type tool makes with a single click) hugs its characters - the layer is exactly as big as the word. Paragraph text (click-drag with the Type tool, and every caption track) is a fixed container the text sits inside, sized independently of its content. Graphics and Titles ▸ Upgrade Caption to Graphicpreserves the caption track's container, so every upgraded clip is paragraph text in a box roughly the width of your safe margins.
Transform operations - Scale, Rotation, the Transform effect, most pop and bounce presets - pivot around the layer's anchor point, which sits in the middle of that oversized box. Unless your caption happens to be dead-center in the container, the visual center of the word and the mathematical center of the layer don't match. A simple 0→100% scale-up makes the word fly in from somewhere off to the side.
The fix, clip by clip
- Select the upgraded clip and open Essential Graphics ▸ Edit, or double-click the text in the Program monitor.
- Right-click the text layer in the Essential Graphics layer list and choose Convert To Point Text. The bounding box snaps down to the size of the actual words. (If the text reflows, fix line breaks now - point text doesn't wrap.)
- Re-center the anchor: with the layer selected, the anchor-point control in the Program monitor (or Effect Controls ▸ Vector Motion ▸ Anchor Point) should sit at the visual center of the word. Hold Ctrl/Cmdwhile dragging the anchor to snap it to the layer's center.
- Re-apply your motion preset. Scale and rotation now pivot from the middle of the text, and the pop behaves the way it did on your test clip.
For position-only animations (slides, drifts) you can skip the conversion - position keyframes don't care about the container. It's scale, rotation and any Transform-based preset that expose the bug.
The part nobody mentions: doing it 200 times
Convert, recenter, retest takes 20-30 seconds per clip once you're fast. A 60-second short upgraded from captions is 50-80 clips - call it half an hour of pure clicking. A 10-minute video is an afternoon. There is no select-all for Convert To Point Text, and copy-pasting attributes across clips pastes keyframes but not the text-layer type, so the work stays stubbornly manual. The full upgrade-caption workflow and its other gotchas are covered here.
That math is why Caption Plug skips the caption-track-and-upgrade pipeline entirely. It transcribes your timeline audio and renders per-word animated caption clipswhere the pop, bounce or highlight is already baked frame-accurately around each word's own center - no containers to convert, no anchors to drag, nothing to fix afterwards.
Quick answers
Why is the text box huge after Upgrade Caption to Graphic?
Caption tracks store text in a fixed paragraph container roughly the width of your safe margins, and the upgrade preserves it. Each graphic clip therefore arrives as paragraph text inside that oversized box rather than point text hugging the words.
How do I convert paragraph text to point text in Premiere Pro?
Select the clip, open Essential Graphics ▸ Edit, right-click the text layer in the layer list and choose Convert To Point Text. The bounding box snaps to the size of the actual words; fix any reflowed line breaks afterwards, since point text doesn't wrap.
Why does my scale animation slide the text sideways?
Scale and rotation pivot around the layer's anchor point, which sits at the center of the inherited container - not the center of the word. Until you convert to point text and re-center the anchor, every transform originates from the wrong spot.