Esports: motion plus overlay type, not Veo inventing brackets

Kling motion. Overlay the score. An esports clip is two objects. The plate is physics: crowd, arena light, a controller, a silent push across a stage. The bracket is a string you already wrote: 2–1, upper final, the seed that would be a lie if a model redrew it. Motion is allowed around the plate. Invented HUD glyphs are not a scoreboard. They are a misspelled result you cannot edit.

Do not ask Veo to typeset a tournament. Generate a clean plate. Type the real line on the locked take.

Silent motion is the plate

Kling 3 Turbo Text to Video is the mess-and-motion engine: fast text-to-video, audio false, 3s through 15s in one-second steps, 1080p ceiling, 720p or 1080p quality, aspects 16:9 / 9:16 / 1:1. Twelve credits is the catalog figure (10 when the discount applies). requires_image is false. You get a silent plate. That is the product.

A silent prompt on a native-audio model still yields a band. A silent prompt here yields silence. Plan the bed after picture lock. Do not prompt "epic soundtrack" at a model whose audio flag is false.

Write the prompt as motion only: camera, lights, what the crowd does. Do not describe a score, a best-of, or a sponsor bug. A generated "team" with the wrong sleeve is a merch problem, not a highlight. Start at 3s, 4s, or 5s unless the beat is actually longer. Vertical if the clip is a Short; 16:9 if it is a YouTube bumper. This slug has three aspects. Pick one.

The bracket is type you wrote

Add timed text overlays to a video burns several different lines of your own text, each with its own start_sec / end_sec. It does not transcribe. It does not listen. You write every line: a question at 0s, a score at 3s, a punchline at 7s. Font, size, colour, and position are set once for the whole overlay set, not per line. Precision is the timestamps you typed.

That is how "2–1" and "Grand Final" stay the strings the desk signed. Seeds change. Copy lock can be later than picture lock. If either is open, do not burn.

One persistent headline — a handle, a "watch live" badge — is a single text overlay, not a timestamp list. Auto-transcribed subtitles of speech are a caption burn, not a score overlay. Do not ask Kling or Veo to typeset standings. A generated "HUD" is pixels you cannot correct without regenerating the motion. Generate a clean plate. Type the real string.

Do not ask Veo to typeset a tournament

Veo is a talking and cinematic row: native audio, short clips (4 / 6 / 8 seconds on Veo 3.1), 4K when the plate needs it. Use it when a caster has a line, or when the shot is a cinematic world. Do not use it as a typesetter. A Veo take that "has the bracket in the shot" is how you ship a 2–1 with a 3–0 hallucination burned into the jumbotron.

Split the jobs:

• Motion B-roll, crowd, abstract stage energy. Kling 3 Turbo. Silent. 3s–15s. 1080p.
• Score, seed, match title, countdown. Timed overlays you wrote. After picture lock.
• A talking talent. A presenter or lipsync row, not Turbo with a mouth you will later VO over.

One generate that tries to be all three is how the score morphs. Crop the keeper, then print the line. Overlaying a 16:9 score onto a feed crop is how the 2–1 lands in the chrome.

FAQ

Can I prompt Veo to "hold the bracket in frame"?

You can. You should not ship it. The model will paint text, not typeset your standings. Overlay the real string on a locked Kling plate. Do not generate the bracket.

Does Kling 3 Turbo Text to Video generate speech if I put the score in quotes?

No. audio is false. Quotes in the prompt are still a silent plate, and any on-screen glyphs are still a generate, not type. Overlay the score.

One overlay or timed lines?

One overlay is one string for the whole clip — a handle, a "live now" badge. Several beats that appear and disappear — score, then MVP, then CTA — need timed overlays. Neither of those is a generate.

Should I overlay before I crop to 9:16?

No. Crop the keeper, then print the spec. Overlaying a 16:9 score line onto a feed crop is how the result sits under the UI.