A product infographic made with AI works best in two layers: the AI makes the picture (the product in a clean scene with deliberate empty space for callouts), and you set the words, numbers and icons in a design tool. On an 8frame board that is a cut-out or a clean still, a scene generated around it with room left for text, a content-preserving upscale, then export to your layout tool. The limit people do not expect: image models redraw every letter on every generation, so specs, measurements and claims drawn by the model come back slightly wrong, and no image model on 8frame outputs vector or editable text.
TL;DR
| Step | Node or model on 8frame | Fixed price | Input |
|---|---|---|---|
| 1. Clean product | Background removal | 12 credits per image | Your product photo |
| 2. Scene with empty space | Nano Banana 2 | 8 / 11 / 17 / 22 credits at 0.5K / 1K / 2K / 4K | Cut-out as reference, layout prompt |
| 2b. Scene, many references | Seedream 5 Lite | 6 credits per image | Up to 14 references |
| 3. Clear a cluttered area | Flux Edit (FLUX.2 pro) | 7 credits for a ~1 MP input, rising with input size | The generated scene |
| 4. Enlarge for zoom | BRIA Increase Resolution | 6 credits, 2x or 4x, up to 8192 x 8192 | The approved scene |
| 5. Text and icons | Your design tool | No credits | The scene as a PNG |
Why the text stays out of the model
An infographic is a picture with facts on it: "12-hour battery", "fits 15-inch laptops", "BPA-free", arrows to features. Every one of those facts is a claim the buyer relies on. Image models do not type; they draw shapes that look like letters, and they draw them again each time. On a single headline word the result can look right. On a row of numbers, a unit, or a small callout, characters swap, double or blur, and a model can quietly change "15" to "16".
There is a second reason. Infographic text changes: a new size, a translation, a price-free variant for a marketplace. Text drawn into a raster image cannot be edited, only regenerated. The image models we measured on 8frame (Nano Banana 2 and Pro, GPT Image 2, Seedream 5) return PNG, and no image model on 8frame outputs vector or SVG. Text set in a design tool can be corrected in seconds and exported in every language from one file.
So the model's job is narrower and easier: make a product picture with a clear composition and blank, evenly lit space where the text will go.
The board, node by node
The live template that comes closest is PRODUCT PLACEMENT on /workflows, which puts a product into a new scene. The infographic recipe adds the layout step and the text handoff.
Step 1: a clean product
Start from your best product photo. If it has a busy background, run background removal (12 credits per image) to get a cut-out. If the photo is already on a plain sweep, skip this; the model can work from it directly.
Check the cut-out edges at 100%: straps, handles and transparent parts (glass, clear plastic) are where cut-outs go wrong. Fix the source photo rather than accepting a ragged edge, because the next step will copy it.
Step 2: the scene, with space reserved
Connect the product as a reference to an image node and write the prompt as a layout, not as a mood:
Product photo of the attached backpack, three-quarter view, standing on a light grey studio floor, placed in the right third of the frame. Left two thirds: plain light grey background, evenly lit, no objects, no text, no shadows crossing it. Soft light from the upper right. Square image.
Name where the product sits and what the empty area looks like. Say "no text" explicitly; without it, some generations add fake labels or captions.
Which model:
- Nano Banana 2 is the default Nano Banana model, priced by size: 8 / 11 / 17 / 22 credits at 0.5K / 1K / 2K / 4K. Draft at 1K, make the final at 4K for marketplace zoom. Nano Banana is also the image generator on 8frame that offers 4:5, which matters if the infographic doubles as an Instagram feed post.
- Seedream 5 Lite is 6 credits per image and takes up to 14 references, the most on 8frame, useful when you attach the product from several angles plus a style reference for a set of infographics.
- Nano Banana Pro is 21 credits at 1K or 2K, 42 at 4K, if you want its look for a hero image.
- GPT Image 2 defaults to high quality at 29 credits square, 20 for other ratios; medium is 8 square, 6 other ratios.
Make four to six candidates and pick for composition: product edges clean, empty area truly empty, light consistent.
Step 3: clear what got in the way
If the chosen scene is right except for a stray prop or a shadow crossing the text area, edit it rather than regenerating. Flux Edit (FLUX.2 pro) is 7 credits for a ~1 MP input and rises with input size; Kontext Max is 13 credits. Ask for one change: "remove the plant on the left, keep everything else the same". Two changes in one edit is two chances for the product to change.
Step 4: upscale with a content-preserving model
For marketplace zoom, enlarge the approved scene before the text goes on. BRIA Increase Resolution is 6 credits, 2x or 4x, content-preserving, up to 8192 x 8192. Topaz is 11 credits per started 24 MP block. Avoid Clarity (13 credits) for product infographics: it is diffusion-based and always enlarges 2x on 8frame, so it can add texture or redraw fine detail on the product.
Step 5: set the text and icons in a design tool
Download the original file (PNG), open it in your layout tool, and add the callouts, numbers, icons and leader lines there. Use the same type styles and positions across every infographic for the product so the gallery reads as one set. Keep the facts in one source document and paste from it, so the infographic says exactly what the product page says.
Put a sticky note on the board next to the final image with the file name of the layout and the date, so the next person knows where the text lives.
Which model for which step
| Need | Model | Price | Note |
|---|---|---|---|
| Product cut-out | Background removal | 12 per image | Image only; there is no video background removal |
| Scene with empty layout area | Nano Banana 2 | 8 / 11 / 17 / 22 at 0.5K / 1K / 2K / 4K | Has a 4:5 option; up to 10 references |
| Set of infographics in one look | Seedream 5 Lite | 6 per image | Up to 14 references |
| Single-change edit | Flux Edit | 7 for ~1 MP, rising with size | One change per edit |
| Edit with more control | Kontext Max | 13 | Can match the input's ratio |
| Enlarge without redrawing | BRIA | 6, 2x or 4x | Content-preserving |
| Enlarge large files | Topaz | 11 per started 24 MP block | Fixed precision path on 8frame |
Where it breaks
- Text drawn by the model. Expect letter and number drift. Keep every word, number and unit out of the prompt's output and set it yourself.
- The product changes. A scene generation can shift colour, proportions or the label. Compare with the real product before the text goes on; see why AI changes product labels.
- Empty space that is not empty. Models fill gaps with props. Describe the empty area precisely and say "no objects".
- A soft scene. If the generated scene looks soft at zoom, check the resolution setting first; why is my AI image blurry walks through the causes.
- Upscaling after the text. Text set before upscaling gets resampled; set text last.
- Main image rules. Infographics are secondary images. Google Merchant Center's image requirements say the main image should be "focused solely on the product itself" and list "post-production text, marketing callouts, or graphics" as not allowed there.
Platform rules that apply
Google Merchant Center's image requirements, checked 2026-11-09:
- Main image: "Your product image must offer an unobstructed, clean view of the item for sale." Post-production text, marketing callouts, graphics, stock photo watermarks and graphic stamps are not allowed on it.
- Additional images: "Any additional views of the product showing highlighted areas can be submitted" through the additional image link attribute.
- Size: at least 500 x 500 pixels, around 1500 x 1500 or above recommended, no larger than 64 megapixels or 16 MB.
- AI: "All images created using generative AI must contain meta data indicating that the image was AI-generated", and you must not remove tags such as the IPTC DigitalSourceType property. Some design tools strip metadata on export, so check the final file.
Shopify says collection pages display best when featured images share "a consistent aspect ratio", and recommends 2048 x 2048 px for square images; images can be up to 5000 x 5000 px or 25 megapixels, under 20 MB.
FAQ
Can AI make a product infographic with text?
It can draw text, but it redraws every letter each time, so numbers and small print drift. Generate the picture with empty space and set the text in a design tool.
Which AI model is best for product infographics?
On 8frame, Nano Banana 2 for single scenes (priced by size, with a 4:5 option) and Seedream 5 Lite for a set that shares one look (6 credits, up to 14 references). Pick for composition and product accuracy; the text is not the model's job.
Can I put an infographic as my main product image?
Google Merchant Center says the main image should show only the product, without post-production text, callouts or graphics. Use infographics as additional images.
Do AI infographic images need a label?
Google Merchant Center requires metadata marking AI-generated images and says not to remove the IPTC DigitalSourceType tag. Check that your export keeps it.
Can 8frame export vector files?
No. The image models we measured return PNG, and no image model on 8frame outputs vector or SVG. Set vector icons and text in your design tool.
Sources
- Google Merchant Center Help, image requirements (main image clean view, no post-production text or callouts; additional views showing highlighted areas; 500 x 500 minimum, around 1500 x 1500 recommended, 64 MP and 16 MB maximum; AI metadata, IPTC DigitalSourceType): support.google.com, checked 2026-11-09
- Shopify Help Center, product media types (consistent aspect ratio; 2048 x 2048 for square; up to 5000 x 5000 or 25 MP, under 20 MB): help.shopify.com, checked 2026-11-09
- 8frame prices, defaults and limits: the 8frame app, checked 2026-11-09
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.