Text in AI video: give exact words an editable layer.
If a title, name or caption must be exact, plan to add it as editable text in a video editor after choosing the background clip. Write the approved copy and its timing separately from the motion prompt. A sign attached to a moving wall is a different task: its lettering must also follow the surface and any objects passing in front of it.
1. Decide whether the text belongs to the screen
Start with the job the words do. A headline stays readable over the picture. A caption follows speech. A shop sign belongs inside the photographed scene. This choice determines what you need to control; asking for all three in one generation prompt leaves those different responsibilities unresolved.
| Text job | What to prepare | What to review |
|---|---|---|
| Screen headline or lower third | Approved copy and an editable title layer | Spelling, line breaks, contrast, position and reading time |
| Speech caption | Checked transcript and timings for the actual audio | Correct words, speaker and synchronization |
| Sign, screen or product label in the scene | Approved artwork and a plan for its surface | Perspective, attachment, occlusion and changing letterforms |
For packaging, use the existing product label and logo workflow. For spoken lines, keep the script and speaker assignments with the dialogue plan. The workflow below focuses on handing off an editable text overlay.
If letters already change shape in the background clip, a new title above them does not repair those pixels. Choose another take, replace the affected area in an editor, or change the shot. Do not treat a correctly spelled headline as evidence that the underlying sign is correct.
2. Copy a text handoff card
Record the exact string before deciding on animation. Preserve capitalization, punctuation, numbers and any required line break. For a wordmark, supply the approved artwork rather than assuming a similar font recreates it. The card is an editor brief, not a syntax that makes a generator create a separate text layer.
TEXT HANDOFF — original planning example Exact copy: Next stop: the coast Copy source/version: approved headline, version 1 Relationship: fixed to the screen, above the background clip Delivery: planned 9:16 export; also review the published player Placement: upper-left open sky; adjust after reviewing the full clip Type: one editable text layer; choose the typeface in the editor In: 00:00:01.000 on the final edit timeline Out: 00:00:04.000 on the final edit timeline Motion: simple appearance, then hold for reading Review: exact copy, contrast, crop, timing and subject overlap Background take: record the accepted clip ID here
These times are illustrative edit decisions, not a claim about any generator's supported duration. Review the actual reading time; shorten the copy or extend its hold when it cannot be read comfortably. Keep a separate card for each delivery crop if the layout changes. See the aspect-ratio guide before assuming a wide layout will survive a vertical crop.
3. Write the background motion and editor notes apart
Choose a reference with space for the words, then watch whether that space remains usable as the clip moves. A prompt requesting open space is a composition target, not a reserved overlay region or a guarantee that the subject will stay clear of it.
Runway's current image-to-video prompting guide, optimized for Gen-4.5, says the input establishes the scene and the prompt should focus on motion. It also recommends a clear source without visual artifacts. This is guidance for preparing the scene; it does not promise exact lettering throughout an animation.
Example A: a destination headline over a landscape
Reference assumption: the coastline occupies the lower-right area and clear sky is already visible at the upper left.
Small waves roll toward the shore while thin clouds drift slowly across the distant horizon. The camera holds the original wide view. The coastline stays in the lower-right part of the composition.
Separate editor noteAdd the approved headline from the handoff card as editable text. Check its upper-left placement throughout the chosen take and in the final vertical crop. Keep the title on screen long enough to read.
Example B: opening-time information over a cafe shot
Reference assumption: a cup and saucer sit on the right of a tabletop, with a clear area on the left. The copy is an invented planning example, not a real business's opening time.
A thin curl of steam rises above the cup. The cup and saucer remain on the right of the table. The camera holds the tabletop view as the steam thins and disperses.
Separate editor notePlace the exact copy “Open at 8” in an editable screen overlay on the left. Set its in and out points after selecting the clip. This is an information title, not lettering painted onto the cup or table.
Both drafts are original, unrendered examples. Match the scene to your real input and inspect the output. Do not paste the editor note into a motion field and assume the video service will perform the later edit.
4. Add the approved copy after accepting the background
In your video editor, place the chosen clip on the timeline and create the text separately. Set the exact copy, appearance, position and duration. For a concrete example, Adobe's Premiere title tutorial demonstrates the Type tool, styling in Properties, extending the title clip and animating its position with keyframes. Use the controls your own editor provides; no particular subscription is required by this planning method.
A screen overlay can stay in one place while the scene moves behind it. A title intended to follow an object needs a different attachment plan. Adobe's After Effects motion-tracking documentation describes applying tracked motion to another layer and using corner pin tracking for perspective changes. That does not make every shot a simple replacement: inspect drift, surface changes and objects crossing in front of the lettering. A graphic floating above a passing hand will not look attached to the surface behind it.
Recheck the timeline after trimming or replacing the background. Keep the editable project as well as the delivered file so a later spelling or timing revision can be made in the text layer.
5. Read the finished export, not only the editor preview
Watch once at the intended viewing size, then pause and scrub the text interval. Read against the approved copy instead of relying on memory. Inspect the entrance, the busiest background moment and the last readable frame. Check the intended player or publishing preview too; its crop and interface may cover a different part of the picture.
| Observed problem | Check | Next action to assess |
|---|---|---|
| Wrong word or punctuation | Compare the text layer with approved copy | Correct the editable string and export again |
| Correct text disappears into the scene | Read over the brightest and busiest background moments | Adjust placement or appearance, then replay the interval |
| The words are gone before you finish reading | Check the final edit's in and out points | Shorten the copy or increase the hold |
| A caption has the wrong words or timing | Compare it with the actual spoken line | Correct the transcript and synchronization in the edit |
| A sign slides away from its wall | Review the surface track, perspective and occlusion | Revise the composite or choose a more suitable shot |
| A moving label changes behind the title | Inspect the original clip without the overlay | Address the background take; changing the title is insufficient |
Original review criteria, not measured failure rates. Scroll wide tables sideways on a small screen, or focus a table region and use the arrow keys.
Save the approved copy version, background take, edit version and a time-stamped observation. “The final word is covered by the person at 00:03” identifies a layout problem more clearly than “AI text is bad.” Choose the next step from the actual failure rather than adding more adjectives to an unrelated motion prompt.
Common text questions
Will quotation marks guarantee readable text in AI video?
No. Check the selected model's current instructions, then inspect the complete clip. A prompt convention is not a guarantee of spelling, letter shapes, typography or consistency. Keep exact copy editable when it needs reliable revision.
Does “no text” belong in every background prompt?
Not automatically. It can conflict with a required sign or product label, and models handle exclusion instructions differently. Consult the negative-prompt guide before reusing an exclusion list. Review the actual background for unwanted lettering even when the brief requests an open surface.
Can Shotmelo add the overlay to my video?
The current Shotmelo preview helps you edit a motion brief, save a local draft and export its text. It does not render video, correct generated lettering, create video text layers or export an edited movie. Use the image-to-video prompt planner for the motion draft and carry the separate copy card to your editor.
Published November 10, 2026. Linked Runway and Adobe documentation was checked on that date. Examples and handoff cards are original planning materials, not rendered tests or claims of a model's accuracy.