A Product-Mockup Generation Workflow

Traditional product mockup creation requires either a 3D renderer (such as Blender with a UV-mapped model), a paid mockup platform subscription, or a set of layered Photoshop templates. Each approach has friction: 3D rendering requires technical skill and rendering time; mockup platforms are limited to their own template library; and Photoshop templates need manual perspective-warp adjustment for every new design. All three approaches also produce results that look like what they are — digitally composited designs rather than genuine photographs of printed products.

AI-based mockup generation works differently. Instead of compositing a flat design onto a photograph, the AI model synthesizes an image of a physical product as it would appear if printed with your design, taking into account the fabric texture of a t-shirt, the curvature of a mug, the surface sheen of a phone case, and the natural lighting of the scene. The result is a photorealistic mockup that looks like an actual product photograph rather than a digital composite, which dramatically improves conversion rates in e-commerce listings.

The Floniks editor supports this through a combination of Image-to-Image nodes (conditioning the generation on both the design and the product base image), ControlNet-style nodes (enforcing the product silhouette and surface geometry), and optional lighting adjustment nodes. Once the pipeline is configured, batching it across 30 design variants takes the same workflow run — only the input design image changes per batch row.

The source design should be on a transparent or solid white background with the printable artwork occupying the maximum canvas size. High-contrast, high-resolution designs (at least 1000x1000 pixels) produce the most faithful results because the model can clearly resolve the design elements before mapping them onto the product surface. Avoid designs with very fine lines thinner than 5 pixels at the source resolution, as these frequently disappear after the surface-mapping step.

The product base image should be a clean photograph or rendering of the product without any design — for example, a plain white t-shirt on a model or mannequin, an empty white coffee mug, or a blank phone case. The base image establishes the lighting, shadows, surface normal direction, and product geometry that the AI will use when synthesizing the design placement. Consistent base images across a product line ensure that all generated mockups have matching background, lighting angle, and product orientation — which matters greatly for e-commerce catalogs where all items should look like they belong to the same photoshoot.

If you do not have a product base image, write a prompt for a Text-to-Image generation node to create one: "plain white cotton crew-neck t-shirt, front view, soft studio light from upper left, white background, no design, photorealistic, lifestyle mannequin." Use this generated base as the input to the mockup node. For consistent results across a catalog, generate the base once and reuse it for every design variant rather than re-generating it per run.

The placement prompt has two jobs: describe the desired position and scale of the design on the product surface, and describe the physical properties of the printing method. For position and scale: "custom graphic centered on chest area, 22cm wide, lower edge at natural chest crease, full-color print, sharp edges, no visible stitching intersecting the design." For printing method: "direct-to-garment print, CMYK ink on cotton, slight ink texture visible on fabric weave, no screen-door halftone pattern."

Different products require different placement keywords. For mugs: "ceramic surface, wraparound design on cylinder, glossy fired glaze over print, slight convex curvature visible, studio soft-box light from left, no handle in frame." For phone cases: "polished hard-shell case, UV print directly on matte plastic, slight edge gloss on case lip, screen and camera cutouts visible, design centered on back panel." For tote bags: "natural cotton canvas tote, screen-print look, visible canvas texture through ink, single-color or two-color ink, center print placement."

The negative prompt should exclude any placeholder graphics, factory logos, or default product markings that the model might invent: "blank, no design, empty product, default logo, manufacturer logo, barcode." Also exclude unrealistic printing artifacts: "floating design, non-physical texture, design not adhering to surface, flat composite look."

Flat design artwork must wrap around curved or folded surfaces correctly to look like a genuine printed product. A mug design must curve with the cylinder. A t-shirt design must follow the fabric drape across the chest. A tote bag design must distort slightly to follow the natural creasing of the canvas. Incorrect wrapping — a flat-looking design pasted onto a curved surface — is the most obvious visual cue that a mockup is artificial.

The strongest mitigation is to include a ControlNet-style surface normal node (if available in the Floniks editor for your selected model) that takes the product base image as the control input. The surface normal map derived from the product base provides the model with precise curvature information for every pixel of the product surface, ensuring the design warps correctly to match the underlying geometry. This is the difference between a design that looks "printed on" versus a design that looks "composited on."

If a surface normal node is not available for the product type, use the following prompt strategy: add explicit perspective and curvature description such as "design follows natural cylinder curvature, left edge slightly foreshortened, right edge slightly foreshortened, consistent perspective with product geometry." This language primes the model to apply curvature even without explicit geometric guidance. Always compare the mockup output against a real product photograph of the same product type to calibrate whether the curvature feels physically plausible.

The highest-leverage use of a product mockup workflow is batching it across an entire design catalog. After validating the workflow on a single design, connect the Design Image Input node to a Batch Input List instead of a single file. Each row in the batch list is a different design file — a different graphic, color scheme, or print — and the workflow generates one mockup per row in a single run.

Output files should be automatically named using a naming convention node that appends the design file name, product type, and run date to the output file name: for example "tshirt-front-[design-name]-2026-06-19.png." This naming structure makes it trivial to match each mockup to its source design in the catalog management system. Connect the output naming node to a bulk download step that packages all generated mockups into a single archive.

For print-on-demand sellers who add new designs frequently, schedule the batch workflow as a recurring template run: whenever new design files are added to a designated input folder, the workflow generates the full mockup set automatically. Credit cost per mockup run is typically equivalent to one standard image generation, so a 50-design batch costs 50 credits — a fraction of the cost and time of equivalent traditional rendering or per-design mockup platform fees.