BLACK FRIDAY

Black Friday: 50% off annual plans + 30% bonus credits on Pro

Get 50% OFF
Ends in --d --:--:--

Seedance 2.5 On-Screen Text and Title Cards: What to Prompt and What to Composite

Nov 22, 2026

Short answer: Treat exact wording in Seedance 2.5 as a post-production job. Leave title-safe space in the prompt for cards and lower thirds, and do not ask the model to spell a headline, logo, or subtitle in the picture. For in-world signs, keep lettering generic or out of focus unless you will composite it later.

Updated: November 22, 2026

How should you handle on-screen text in Seedance 2.5?

Separate two jobs that are easy to mix up. One job is the moving picture: subject, action, camera, light, and sound. The other job is exact type: a title card, a lower third, a subtitle, or a logo lockup that must match a brand file. For Seedance 2.5, plan the picture first and add the final wording after you export, unless you are deliberately creating a loose background graphic that no one has to read.

This is a workflow choice, not a claim that every generated frame will fail. Exact spelling, kerning, and brand fonts are still a typesetting task. If you need a letter-perfect headline, keep the frame clean and composite the type in an editor you already use. A related starting point for the first shot is Seedance 2.5 first clip from one photo; if you already know you need a title, reserve the space before you generate.

Which kinds of text should be composited instead of generated?

Use the table as a default route, then adjust if the type is only atmospheric and does not have to be legible.

Text typeRecommended route
Opening or closing title cardsLeave a clean title-safe area in the prompt; composite the final title after export
Lower thirds and name titlesKeep the lower band simple and without typography; add the name plate in post
Subtitles or captionsDo not ask the model to paint readable subtitles into the frame; add them after export
In-world signs, posters, and storefrontsKeep lettering generic, blank, or out of focus unless you will replace it later
Packaging and product labelsOut of scope for this guide; treat readable packaging as a separate composite or reshoot problem

Do not treat these routes as a measured typography test. They are the safer path when the wording has to be exact. If a background poster only needs to feel like a poster, a soft or blank graphic is enough. If a client has to read a slogan, composite it.

How do you prompt title-safe space for cards and lower thirds?

Say what should stay empty, not only what the scene contains. Name the region (top title band, lower third, or a held title card), keep that region visually simple, and state that the model should not render a headline, logo, or caption in the picture. Then describe the shot as you usually would: subject, action, camera, and light. A general prompt walkthrough is in the Seedance 2.5 prompt guide, and Seedance 2.5 text to video covers the text-to-video path when you are not starting from a still.

A locked camera often makes typography easier to composite later, because the empty band stays in one place. If the shot must move, see Seedance 2.5 static camera fixes for ways to keep the frame stable, then still add the type after export.

The following blocks are starting prompts, not a guarantee of a specific frame. Change the subject and setting to match your brief, and keep the “no readable type” instruction.

Title-card plate.

A wide, cinematic city street at blue hour, wet asphalt reflections, empty sidewalk in the foreground. Leave a clean, dark, uncluttered band across the lower fifth of the frame for a title to be added later. No readable text, no logos, no captions, no subtitles. Static camera, 16:9.

Lower-third interview shot.

Medium shot of a presenter in a bright studio, looking to camera, soft key light and a simple background. Keep the lower-left third of the frame visually plain and without signs or typography so a name title can be composited later. Do not render any headline, name, or caption in the picture. Locked camera, 16:9.

What should you do with in-world signs and posters?

Scene graphics are different from titles you will read in the edit. A storefront, street sign, or wall poster can stay in the world, but readable lettering is still a spelling problem. Prompt the object’s shape, material, and placement, and keep the lettering blank, generic, or too soft to read unless you plan to replace it.

Storefront with a blank sign.

A neighborhood bakery storefront in late afternoon light. A painted hanging sign and a window poster are visible as shapes, but the lettering is blank or out of focus and cannot be read. Do not generate logos or written words. Slow push-in, 16:9.

If the sign itself is the product, treat the readable art as a later composite: generate the hand, the wall, and the lighting, then place the real artwork on the sign in post. That is the same idea as reserving a title-safe band, applied to an object in the set. When several shots share one location, keep the sign placement consistent with the rest of the set; a reference-led workflow is in Seedance 2.5 workflow from reference images.

Looping clips make baked-in wording even harder to live with, because a typo or a drifting letter will repeat. If you are building a loop, leave the type out and add it after you lock the motion. See Seedance 2.5 loop video prompts for the loop itself.

How do you review a frame before adding the final type?

Export the clip, then check still frames, not only the moving preview. Confirm three things before you design the typography:

  1. The reserved band or sign area is still clear enough on the frames where the type must appear. If a hand, prop, or highlight has moved into that area, regenerate with a stricter empty-region instruction or pick a different hold frame.
  2. There is no accidental lettering elsewhere in the shot: walls, shirts, screens, or packaging. If stray characters appeared, do not use them as the final title. Prompt again with a stronger “no readable text” line, or mask and replace the area.
  3. The frame has contrast and space for the actual type. A busy background may still work if you add a temporary backing plate or a simple bar in post; do not ask the model to invent that plate with words already on it.

Only after those checks should you set the final font, size, and wording. If the title has to survive a translation later, compositing also keeps the picture reusable when the language changes.

What are the common questions about on-screen text in Seedance 2.5?

Can Seedance 2.5 render an exact, brand-ready title in the picture?

Do not plan on it for wording that has to match a script or a brand file. Leave space and composite the title after export.

Is this the same as adding subtitles?

No. Subtitles and captions also belong in post if they must be accurate. This guide is about title cards, lower thirds, and in-world signs, not a subtitle-file workflow.

Should I put “no text” in a negative prompt, or just avoid asking for words?

Do both when the shot matters: say that the reserved area stays empty, and say that the picture should not contain a headline, logo, caption, or readable sign. Naming the region is more useful than a generic “no text” line alone.

When is a blank sign better than generated letters?

Use a blank or out-of-focus sign whenever the audience, a client, or a later language version needs to read a specific name or slogan. Generated letters are only a fallback when the graphic is atmospheric and no one has to read it.

Do I need a new generation if I only want to change the title?

Not if the type is composited. Change the title in the edit and keep the same picture. That is the main practical reason to leave the wording out of the model’s frame.

Try Seedance 2.5.

Seedance Team

Seedance Team