LED signs: the type must be true, the motion is later
Lock type. Then I2V if needed.
Lock type. Then I2V if needed. An LED sign is a machine whose product is letters. If the demo reads OPNE at thumbnail size, the product failed in the ad. Cinematic chrome on melted glyphs is a demonstration that the sign does not work.
Do not ask a video model to render the message. Typeset the copy on a still built for lettering, or stamp the line onto a clean cabinet after the clip exists. Those are the only two honest paths. Motion is a later bill.
The letters are the product
Buyers are buying a string that will sit over a parking lot: a name, hours, OPEN, a price they will be held to. If a character is wrong, the cabinet is wrong. Brand names and bilingual lines are the failure modes. A wet lot in the background is survivable. The matrix is not.
So the still you lock is either:
- a photograph of a real cabinet with a real, proofread message on the face, or
- a generated or photographed clean cabinet — no message — that you will letter afterward.
Ideogram V4 is the stills row when the message has to be generated: six credits, text rendering, posters and logos listed, aspect ratios 1:1, 16:9, 9:16, 4:3, 3:4. Quote the string. "OPEN UNTIL 9" is the brief, not "a night sign with words." Case, alignment, how many lines. Proof at 100%. A wrong digit is a different offer.
If the copy is already on a cabinet you photographed, do not regenerate the lockup. Keep the photograph. Ideogram is for the graphic you do not have yet. Overlay conditions in the editor — do not ask a poster engine for three lines of small print.
Lock the still, then buy motion
Image-to-video starts from the frame you already own. Frame one is the file. The prompt describes only the camera: a slow push to the face, a hold, a slight parallax on a cabinet that is already correct. Do not re-describe the message. If you find yourself writing "a dramatic LED wall that says SUMMER SALE," you have left the still and gone back to a video model as typesetter.
On-screen text in generated video is why the order is this order: two hundred frames is two hundred chances for a letter to flip. A still with correct type is not immune to drift, but it starts from a correct frame one.
Duration is a dial. A three-second hold plus a caption is an ad. An eight-second invented storefront is a mood board you will still have to letter. If you need a line you can edit without regenerating the cabinet, add a text overlay burns one authored string. Do not overlay and leave generated copy on the matrix.
What motion is allowed to do
Move the camera. Hold the type. Night can be in the plate if you shot night. Do not prompt a video model to "turn the sign on" if that means restyping the face. A clean cabinet plus overlay is the switch-on you can correct.
Daytime cabinets still have to read. Do not chase a wet-night cinematic take to hide a misspelled still. The hardware can be cinematic. The message cannot be approximate. Ideogram for the still that already contains correct type. Image-to-video after you would actually post the PNG. Overlay for the line you need to change on Friday.
FAQ
Can I animate a photo of a real LED face?
Yes. That is the best path. The type is already correct. Image-to-video from that photo, motion-only prompt, overlay only if you need a claim the face does not already carry.
Why not generate the copy in Ideogram, then ask a video model to keep it?
You can lock the still and buy motion, and small lettering will still drift under a moving camera. If the message is the product, overlay on a clean face, or keep the camera off the tiniest line and show the real photo as a cutaway. Do not treat I2V as a typesetting guarantee on six-pixel copy.
Does overlay look fake on a moving matrix?
It looks like a caption if you treat it as a caption — a claim, hours. It looks wrong if you fake perspective on a curved cabinet with a top-center burn-in. For that shot, photograph a real face. Overlay is for words you cannot let the model invent.
Is Ideogram the only still model that can letter?
No. Pick a still row that names text rendering, proofread, then decide motion or overlay. Do not pick a video row because the cabinet looked cinematic in a demo reel. Six credits to spell. Then a video row for the push.