Shelter cutaways are Qwen labeled sections at 2K, 4cr

Qwen Image 3 2K/4cr labels the FEMA cutaway; Veo invents a storm that is not the rated room, and a talking meteorologist is the wrong row.

A storm-shelter listing that informs is a board: wall, roof, anchoring, the section names the builder already prints. Those strings are the product. A cinematic tornado with a constructed meteorologist is a weather reel, not the rated room. Letter the cutaway. Leave the storm off the generate.

The FEMA cutaway is a labeled board

Procurement and a homeowner pause on labels they can check against a drawing: wall, roof, anchor, door, the exact strings on the cutaway you will ship. Put that board on a still. Proof it at 100%. Do not bury it in a video prompt and hope the glyphs hold for eight seconds.

A cutaway is named parts, not a vibe. If the drawing has five headings, the board has five labels. Extra tiles “to see options” are cousin documentation.

Do not brief “cinematic tornado shelter, family inside, debris flying.” That is a storm you do not stage, and a family you did not cast. Brief the headings. If a rating line must stay verbatim, say so and inspect the still. Paraphrasing a listed rating into a stronger one is a claim.

A constructed meteorologist walking “what you should do when the siren sounds” is a mouth this leftover cannot carry. Keep people out of the plate. Empty section. Readable type.

Home builders own new construction. This leftover is the shelter cutaway. Do not rewrite that audience as a FEMA board.

Qwen Image 3 at 2K, 4 credits

Qwen Image 3 Text to Image is the stills row when the letters have to read. Catalog facts: Qwen’s third-generation image model, strong prompt adherence, complex text rendering in Chinese and English, intelligent prompt rewriting on by default, 4 credits, 2K max, qualities 1K and 2K. No audio. No duration. Aspects include 1:1, 16:9, 9:16, and 21:9.

Write the exact strings. If a section name, a rating line, or a “safe room” lockup must stay verbatim, say so in the prompt and inspect the still. Rewriting is a help for scene language, not a licence to paraphrase a listed rating into a neighbour’s. Set 1K while the words are in draft. Set 2K when the cutaway is the cutaway. There is no 4K chip on this row.

Qwen is not a storm engine. It typesets a lockup. Photograph a real drawing when you have one. Generate a labeled section when the brief is the word on a plate. Proof the letters before you buy seconds. Do not ask Qwen for a photoreal meteorologist pointing at radar. That is a host. The SKU is the board.

Generate an image from text is the named agent job: you describe the still, the agent asks about style, aspect ratio, and count, then picks a model. Cost is priced per model, shown before you confirm. Name Qwen. A bare “make a storm shelter ad” will land on a portrait row. Quote the sections. Scout at 1K. Keep at 2K when someone will zoom. Proof at 100%.

A video model is a bad typography engine. Do the lock here. Approve the characters. Then consider whether any still even needs to move.

Veo invents a storm that is not the room

Veo 3.1 is native audio, official 4K tier, 4 / 6 / 8 second clips, 40 credits. That is a scene. Text-to-video of “FEMA storm shelter, cinematic tornado, meteorologist explaining wind speed” will decide a cousin room and a presenter who never stood in your shop. 4K of the wrong storm is still the wrong storm.

A video model redraws letterforms every frame. Do not I2V a blank bunker and hope the model stamps section names in flight. Do not prompt debris through a door you did not photograph. Do not prompt a talking meteorologist. This leftover is the board, not a mouth.

“Professional weather presenter, clean studio, what you should do in this storm” on Veo is how you spend the expensive talking meter on a persona. Hold the cutaway. A real named builder on a different day if someone must speak. Do not ask Veo to typeset rating strings. You will get a debris reel you cannot ship.

Overlay ratings after lock; skip the meteorologist

Rating numbers, a county you actually serve, a CTA — those strings move. Generated type inside a video is a misspelled rating you cannot edit without a new generate. Burn the string after picture lock.

Add a text overlay burns one authored line for the whole clip: hook, headline, CTA, the rating, the section that is the hook. You write the line. Position is top, center, or bottom. It does not transcribe. A mute cutaway has nothing to listen to.

Crop to the ratio you will ship before you overlay. Do not overlay a rating you cannot point at a drawing for. If the printed line is already sharp on the Qwen still, you may not need an overlay.

Most listing weeks do not need the cutaway to spin. The board is the product. If a short orbit helps, photograph the real room, then consider still-to-motion on that file. Prompt the camera. Do not prompt the storm. Keep lips off the board. The board is the expert.

FAQ

Can I put the section names in a Veo prompt so the cutaway typesets itself?

No. Veo will invent a storm and a room. Letter the sections on Qwen Image 3 at 4 credits, 2K, and proof the characters. Motion, if any, is a later still-to-motion pass of an approved board — not a flythrough of debris.

Should the agent pick the image model for me?

Generate an image from text will pick a fitting row unless you name one. For a labeled cutaway, name Qwen. A portrait model will give you a mood of weather, not a board you can proof.

When do I overlay instead of baking type into the Qwen still?

Overlay when the line may change — CTA, a county, a rating rewrite. Bake only when the type is the scene and you will not edit it. Do not chain a stamp on top of baked-in type or you double-print the heading.

Why not Veo 3.1 on every shelter because it looks more “brand”?

Veo is the dialogue and 4K row. A talking generate of a meteorologist is the wrong object. Qwen at 2K, 4 credits, is the cutaway job: labeled sections, no storm, no mouth.