Font foundries: Ideogram owns the letters in the frame
Do not Veo the type specimen.
Do not Veo the type specimen.
A foundry's job on a still is spelling. The sample has to be this family, this weight, this string — a pangram, a wordmark, a three-line showing — readable at thumbnail and at 100%. That is not a cinematic generate. Veo 3.1 is 4 / 6 / 8 seconds of scene with native audio and a 4K tier. It will give you atmosphere. It will not set your type.
The catalog model for words inside the pixels is Ideogram V4: 6 credits, text-to-image, features listed as text rendering and posters & logos. The longer argument is Ideogram owns words in the frame. This page is only the foundry case: specimen, not trailer.
A specimen is a spelling job
The type is the picture. A viewer who screenshots the post still has to read the family name. That is the same test as a poster, applied to a foundry catalog:
- A pangram in one weight, one size, on a flat field.
- A three-line showing: display, subhead, caption.
- A glyph set or figure set you actually ship, not a "font-shaped texture."
- A pair of weights compared in one frame.
It is not a 16:9 walk through a sunlit studio with "Garamond" melting on a shop window. If nobody has to read the letters, you are not selling a font. You are selling a mood. Mood is a different row.
Quote the string. Specify case and how many lines. Proof at 100%. A specimen with one wrong letter is a re-render, not a vibe.
Do not send the alphabet to Veo
Veo is a video model. Video models treat lettering as a texture that happens to sit in the scene. Over eight seconds that texture crawls, respells, and grows a serif you do not license. You cannot sell a trial from a clip that cannot hold the word "Renaissance" for six seconds.
The split is upstream:
| Brief | Row |
|---|---|
| Readable letters in a still | Ideogram V4 |
| Photoreal desk, no type to read | A plate model, type composited later |
| Motion on an already-locked specimen | Image-to-video on that still |
| A cinematic B-roll with no specimen | Veo, and do not put the pangram in the prompt |
If you already have the outlines, do not regenerate the outlines. Set the real font in a layout tool and shoot or generate the scene around a raster of that setting. Ideogram is for lettering that does not exist as a file yet — a campaign line, a mock specimen, a social card. A foundry that owns the files should composite, not roll dice on a video model.
Lock the still before any motion
When a specimen needs to move — a slow push, paper fiber, a page turn — animate the approved still. Do not prompt Veo to "hold this type." Text-to-image is the picker; Ideogram V4 is the model choice inside it. Image-to-video is a later door, and it only works if frame one already spells.
Do not ask the motion model to add a second line of copy. The extra string is where the miss lands. If the showing needs a new line, go back to the still.
Foundry feeds die in two ways. One: a beautiful misspelling. Two: a correct spelling that only exists inside an 8-second clip you cannot pause as a catalog image. Ship the still. Motion is optional. Veo is not the specimen engine.
FAQ
Is Ideogram V4 a font renderer I can license from?
No. It is an image model with a text-rendering feature. It will not emit an OTF. Use it for pictures of lettering — cards, posters, mock showings — not as a substitute for your outlines.
Why not overlay the real font on a Veo plate?
Do that when you already have the files and the string must match the retail font to the glyph. Overlay is control. Ideogram is for lettering that is the illustration. Veo-then-overlay is still better than Veo-as-typesetter.
Can I use Ideogram to show a whole glyph set?
Short-to-medium on-image copy is the job. A newspaper page of 200 glyphs is a layout tool, not a poster engine. Grid a few weights. Do not ask one generate to be the PDF specimen book.
What if I only need a moving social teaser?
Generate the specimen as a still on Ideogram V4. Proof the spelling. Then animate that file. The teaser is motion on a lock, not a Veo take that is supposed to remember your capital G.