Guides

    Seed packets: the packet words are the product

    Ideogram.

    Versely Team4 min read

    Ideogram. A seed packet is not a pretty envelope. It is a labelled SKU: variety name, days to maturity, lot, the illustration the gardener already memorised from the rack. The words on the packet are the product. A generate that restyles the type has sold a different cultivar.

    Do not start in a video model. The pack is a still. The still is a contract. Ideogram V4 is the stills row for that contract.

    Packet words are not a style

    Most image models will give you a garden. Ideogram will give you a sign. That is a job split, not a quality ranking. Mixing them is how you spend a day regenerating a variety name a specialist would have spelled on the second try.

    Catalog features: text-to-image, text rendering, posters and logos. Six credits, flat. No duration. No video resolution. Aspect ratios: 1:1, 16:9, 9:16, 4:3, 3:4. Generate the ratio you will ship.

    The type is inside the pixels. A viewer who screenshots the packet still gets the variety name. That is the job for a rack card or a pack shot. It is not a ninety-word growing guide. Ideogram is strong on short-to-medium on-image copy — a headline, a variety, a three-word stamp.

    If you already have a photographed packet, skip the generate. Photograph is ground truth. Ideogram is for the lockup that does not exist yet. Treat the generated type the same way you would treat a print proof. Once it is in the reference set, every subsequent shot will faithfully reproduce whatever typo made it through.

    Six credits to typeset, then motion

    Approval means a screenshot still reads. Zoom to the variety line. If a letter is wrong, reroll the still, not a clip. Six credits is cheap enough to be strict. A video credit is not.

    requires_image is false on this row. You do not have to upload a packet to generate one. If you already have a packet and need a change, an edit row may be kinder than a new V4 roll that restyles the whole lockup.

    No audio, because there is no audio on an image. If the brief says “the packet talks,” you need a voice file after the packet exists as pixels. V4 will not speak. It will typeset. Then, and only then, buy motion.

    Reference-locked product video, not a cousin cultivar

    The failure mode of AI product video is not ugly footage. It is a beautiful shot of something that is almost your packet: the variety re-typeset, the illustration half a shade off. Nobody ships that ad, so the generation was worth nothing.

    The AI product video generator is the tool shape for that job. It runs on reference-to-video models: two or three clean angles of the real object tell the model what must not change — the label, the illustration, the exact colour — while everything else in the frame is open. That is a different failure mode than a text-prompted packet, where “a seed packet like mine” can quietly become a packet that is not.

    Photograph the packet: even light, plain background, three or four angles including the front of pack. Describe the scene, not the product. Check the variety name at full size before you keep a take. Do not generate the packet inside the video model from a paragraph. Describing a packet you already have is how V4’s six credits get wasted twice.

    Proofread the variety name before any orbit

    Freeze the frame where the lockup is most visible. Crop to the type at 100%. Compare word-for-word against the canonical still. A garden that is not on the packet is b-roll — keep the cultivar out of that generate. Generate 9:16 first. Do not generate 16:9 and crop the variety name out of the poster.

    FAQ

    Does Ideogram V4 output video of the packet?

    No. Content type is image. Category is text-to-image. Six credits. Animate after you approve the still.

    How many product photos does a labelled packet need?

    Three or four is usually enough: front, three-quarter, and any side that carries branding. Clean, evenly lit and in focus matters far more than how many you send.

    Will my variety name stay readable in product video?

    Reference-to-video models hold packaging far better than a text prompt ever will, but fine print is still the weakest point. Keep the packet large in frame and check every take at full size.

    Can I place my packet in a garden I have never shot?

    Yes — that is the point of the product-video format. The object comes from your references; the scene around it is generated. The packet still has to be the packet you sell.