Imagen 4 is the best choice when you need to create product photos from a text description alone — no reference photo required. It generates photorealistic images with exceptionally accurate lighting, material rendering, and surface detail. The three-tier quality system (Fast at 3 credits, Standard at 6, Ultra at 9) lets you iterate cheaply on concepts with Fast, then render the final hero shot at Ultra for near-studio quality. Unlike FLUX Kontext or Nano Banana Pro, Imagen 4 is text-to-image only — it imagines the product from your description rather than editing an existing photo.
Add an Image node and select Imagen 4. Set the Version to "fast" (3 credits per image). Write your initial product description and generate 3-5 variants to explore different compositions, angles, and staging ideas. At only 3 credits each, Fast tier is the cheapest way to find the right creative direction before committing to a high-quality render. The visual quality is lower, but the composition, colors, and overall concept are representative of what Standard and Ultra will produce.
Imagen 4 excels at rendering specific materials — frosted glass, brushed aluminum, matte ceramic, natural leather grain. The more precise your material description, the more photorealistic the result. Instead of "a nice bottle," write "a frosted glass bottle with a gold anodized aluminum cap, catching soft diffused light from the left, subtle caustic reflections on the marble surface beneath." Include lighting direction (side, overhead, backlit), surface material (marble, wood, concrete), and atmosphere (studio, lifestyle, outdoor).
Once you have a composition you like from Fast iterations, copy the prompt and switch Version to "ultra" (9 credits). Ultra tier adds significant detail in material surfaces — you can see individual leather grain, micro-textures in brushed metal, and accurate caustic patterns in glass. This is the tier for product listing hero images, print catalogs, and any context where the image will be viewed at full size. The visual difference between Standard (6 credits) and Ultra is noticeable in close-up crops but subtle at thumbnail size — save Ultra for images that will be zoomed into.
Imagen 4 supports 1:1, 16:9, 4:3, 3:4, and 9:16. For Amazon and Shopify product listings, 1:1 is standard. For Instagram product showcases, 4:3 maximizes feed real estate. For website hero banners, 16:9 is conventional. Note that Imagen 4 generates one image per run (no batch generation), so plan to run multiple generations if you need a product shot series.
Beauty product — the material specificity ("frosted glass with gold cap") is what makes Imagen 4 shine. Vague descriptions like "a nice serum bottle" produce generic results. The "marble surface with water droplets" adds lifestyle context that elevates the shot beyond a plain white background.
Professional product photography of a luxury skincare serum bottle, frosted glass with gold cap, on a marble surface with fresh green leaves and water droplets, soft diffused studio lighting, beauty brand aesthetic, 1:1
Tech flat lay — "overhead shot" and "natural daylight from a window" give Imagen 4 specific lighting instructions. Without these, the model defaults to generic studio lighting. The "matte white finish" material description ensures the earbuds render with the correct surface texture.
Flat lay photography of wireless earbuds in charging case, matte white finish, arranged on a clean desk with a notebook and espresso cup, overhead shot, natural daylight from a window, tech product lifestyle photography, 4:3
Imagen 4 has three Version tiers: fast (3 credits), standard (6 credits), ultra (9 credits). Use fast for ideation (3-5 concept explorations), standard for social media posts, and ultra for hero product images and print.
This model is text-to-image only — it cannot edit existing photos. If you have a real product photo and want to change backgrounds or staging, use FLUX Kontext instead.
Imagen 4 handles glass, metal, ceramic, fabric, and leather textures with exceptional accuracy. Always name the specific material and finish (matte, glossy, brushed, frosted) for best results.
For a consistent product series (e.g., 5 products in the same brand), keep the identical lighting and surface description across all prompts and only change the product itself.
Imagen 4 Ultra produces the most photorealistic text-to-image product shots on Martini — material surfaces, lighting, and reflections are nearly indistinguishable from real studio photography. The trade-off is that it only generates one image per run and only works from text (no image input). For editing existing product photos, use FLUX Kontext (Pro at 6 credits, Max at 12). For highest-resolution output (4K), use Nano Banana Pro. For the fastest concept iteration at lowest cost, Imagen 4 Fast at 3 credits per image is unmatched.
Connect Imagen 4 with other AI models on Martini's infinite canvas. No GPU required — start free.
Get Started FreeBlack Forest Labs
FLUX Kontext is the best model for product photography that requires placing your actual product into new scenes. Unlike Imagen 4 (which imagines a product from text) or Nano Banana Pro (which composites from multiple references), FLUX Kontext takes your real product photo and re-contextualizes it — swapping backgrounds, adjusting lighting, and placing the product in lifestyle scenes while preserving the exact appearance of labels, colors, and textures. Two quality tiers are available: Pro (6 credits/image) for standard shots and Max (12 credits/image) for hero images requiring maximum detail fidelity. You can generate 1-4 images per run across 9 aspect ratios.
View guideNano Banana Pro is the highest-resolution image model on Martini, offering 1K, 2K, and 4K output tiers. At 4K, product textures — leather grain, brushed metal, fabric weave — render with enough detail for print catalogs and large-format displays. Unlike Imagen 4 (text-only), Nano Banana Pro supports up to 8 reference images, making it powerful for creating product shots that maintain visual consistency with existing brand assets or reference photos.
View guide