Test-prep: the whiteboard is the product

Qwen/Ideogram for the slide. Not Veo writing math. A SAT short that cannot be paused and read is a classroom mood. "Triangle inequality in four beats" is a page: headline, the inequality in quotes, three worked lines, a source. Veo 3.1 will give you a cinematic whiteboard with number-shaped chalk that almost looks like a proof. Almost is a fail. The student is there to read the line.

Route the still to a lettering or slide model. Then, if you even need motion, move the slide. Do not ask a video model to invent the arithmetic.

The equation is a string you can proof

The deliverable is whether every symbol is the symbol you specified. Thumbnail vibe is not the test. Someone will screenshot the frame and send it to a parent. If the exponent is texture, you shipped noise.

That job is named in Qwen Image 3 Pro when the slide is the product: newspaper-grade text, columns, a deck you will proof character by character. The published family page is Qwen Image 3 Text to Image. The 4-credit text-to-image row is the layout scout. The Pro text-to-image row is 8 credits, up to 2K, text rendering listed — that is the keep. Do not pay Pro for a moodboard of a calculator.

Eight to twelve words that must read at thumbnail size. "SAT Math — no calculator." A module stamp. That is Ideogram V4: 6 credits, posters and logos listed. Quote the string. Do not ask a photoreal workhorse to spell the test name. Ideogram owns words in the frame is the lettering split. Ask it for a ninety-word derivation and you have routed a page to a poster engine.

If you cannot say which of those two you are making, the brief is unwritten.

Veo will give you a classroom, not a SAT line

Veo 3.1 is 4 / 6 / 8 second cinematic video, native audio, official 4K tier. It is a scene generator. A scene generator asked for "clean SAT whiteboard, 9:16, formula on screen" will give you a pretty classroom and formula-shaped light. It will not hold "If a + b > c, then…" as a string you can diff against a copy deck.

Paying for motion before the slide exists is how you spend video credits to discover the inequality is misspelled. Lock the page. Proof at 100%. Then decide: still carousel, slideshow, or a gentle push on a locked frame.

If type will be overlaid later, generate an empty plate. A model that renders words you will cover is paying twice. The slideshow maker exists because a lot of "whiteboards" should never have been generated type. Use Qwen when the slide is the image. Use Ideogram when the stamp is the image. Use Veo for neither.

Text-to-image is the door. Quote every string. "Triangle inequality — three checks." is a string. "A headline about SAT math" is how you get a plausible headline you cannot ship. Run the 4-credit pass first. If columns hold, rerun on Pro. If they do not, tighten the layout block before you pay Pro.

Scout the page, then maybe push the camera

A test-prep account can ship a still. If you want motion, image-to-video on the locked slide is a camera move, not a new page. Do not prompt Veo to "write the formula as it speaks." Writing is typesetting over time. Typeset, then move.

Do not put readable body copy in a lifestyle scene of a desk. Generate the slide, then composite. The test is pause-at-second-two: if the frame is not a usable whiteboard without the motion, the motion was hiding a missing page. Repair is an edit when one string is wrong. It is a regenerate when the proof is texture.

FAQ

Should every SAT still be Qwen Image 3 Pro?

No. Scout on the 4-credit Qwen Image 3 row. Keep on the Pro variant when someone will proof a derivation. Title cards with a handful of words are Ideogram.

When is Ideogram the right row?

When eight to twelve words must read at thumbnail size — a section stamp, a "no calculator" card. A worked proof is a page. Send pages to Qwen.

Can I skip the still and let Veo typeset the formula?

No. Veo 3.1 is a short cinematic clip with native audio. It is not a typesetter. You will get a classroom and formula-shaped chalk.

Why not overlay every number in the editor?

Overlay when the plate is a photograph of a real board you shot. Generate the type when the type is the plate. Mixing both on one string is two sources of truth.