A Lookbook-Generation Workflow

A lookbook is not a collection of isolated images — it is an argument. Every spread says: this model, this brand, this season, this feeling. When each image is generated independently, model face, build, and skin tone drift between shots, styling cues contradict one another, and the lighting language shifts from editorial to commercial mid-spread. The resulting PDF does not read as a unified collection; it reads as a mood board sampled from five different photographers working on five different projects.

The solution is to encode the brand decisions once and replay them across every image in the collection. A workflow in the Floniks editor captures those decisions as locked node configurations: the character reference that defines the model, the background scene library, the color-grade LUT, the aspect ratio for each export format. When a creative director approves the workflow on the first look, every subsequent look inherits the same decisions automatically. The creative effort concentrates in a single approval gate rather than being re-litigated image by image.

The production gain is substantial. A traditional e-commerce lookbook shoot involves booking a model, a photographer, a stylist, a location or studio, and a post-production retoucher. A Floniks lookbook workflow replaces most of that pipeline with a single session in the editor, preserving budget for the physical samples and the marketing spend that actually drives revenue. For seasonal drops where speed to market is critical, the workflow approach can compress a two-week production cycle into a single afternoon.

The character reference is the most important configuration in the entire lookbook workflow. It defines the model who appears on every page. In the Floniks editor, add a Character Reference node and upload a high-quality reference photograph — ideally a clean, well-lit portrait against a neutral background. The node extracts the identity features that will be carried forward through every generation: facial structure, skin tone, hair color and length, approximate build.

In the node settings, set the identity strength to a value between 0.75 and 0.85. Lower values give the model node more creative latitude to adapt to each outfit and scene context, which tends to produce more natural-looking poses. Higher values lock the identity more rigidly, which is useful when the brand has contracted with a specific talent and needs to maintain recognizable likeness. For most fashion lookbooks, 0.80 is a reliable starting point.

Connect the Character Reference output to every downstream Image Generation node in the workflow. Each generation node receives the identity embedding alongside the outfit prompt, so the same model appears in every look: "full-length fashion photograph, [OUTFIT_DESCRIPTOR], model facing three-quarter left, relaxed confident pose, bright airy studio, white seamless background, Vogue editorial quality." The [OUTFIT_DESCRIPTOR] is the only variable that changes between looks. This architecture means a ten-look lookbook requires ten prompt fragments — one per outfit — not ten complete prompt engineering sessions.

Each look in the lookbook is a branch off the character reference node. In the editor, the simplest architecture uses a fan-out pattern: the Character Reference node connects to ten parallel Image Generation nodes, one per look. Each generation node receives the same identity embedding but a different outfit prompt.

For looks shot in a studio environment, the background is synthesized directly in the generation prompt and no separate background node is needed. For looks that require specific environmental contexts — an outdoor terrace for a summer dress, a gallery interior for an evening gown — insert a Background Scene node between the generation and the composite. The Background Scene node produces a photorealistic environment image that the Composite node then blends with the generated model image using an alpha channel from a Background Removal node.

The scene prompts for an outdoor terrace might read: "sun-drenched rooftop terrace, Mediterranean architecture, potted olive trees, golden afternoon light, soft shadows, fashion editorial backdrop." Keep the scene prompts consistent in tone — all aspirational, all high production value, all matching the seasonal palette — so that when the spreads are laid side by side, the environmental variety feels curated rather than random. Archive the approved scene prompts in the workflow notes field so next season the creative team can start from the same vocabulary rather than rebuilding from scratch.

Individual image quality is necessary but not sufficient for a lookbook. What elevates a collection from good images to a recognizable brand document is the consistent chromatic language across every spread. In Floniks, a Color Grade node placed downstream of every generation branch handles this automatically.

Build the color grade to match the season brief. A spring collection might use a warm, slightly desaturated treatment with lifted shadows and a peachy highlight tone: "soft warm grade, lifted shadows to 15 IRE, warm highlights 3200K tone, matte finish, reduced saturation on blues and greens, golden skin tones." An autumn collection might reverse this: "rich cool grade, deep shadows, desaturated midtones, burgundy and ochre accent retention." Store the grade parameters in a shared Color Config node so that all ten grade nodes in the workflow read from the same configuration. Changing the seasonal grade once updates all ten looks simultaneously — a significant advantage when the creative director requests a grade revision after reviewing the first draft.

After grading, pass each image through a Sharpening and Detail node that enhances fabric texture without affecting skin tones. The prompt: "enhance fabric weave and texture, preserve skin tone smoothness, editorial sharpening on garment details, no over-sharpening on face." The final step before composite is a Crop node that applies the golden-ratio crop favored in fashion editorial, placing the model at the left or right third of the frame with space for headline typography on the opposite side.

A professional lookbook ships in at least three formats: a high-resolution PDF for print, a compressed PDF for email and web download, and individual JPEGs for social media distribution. The Floniks export branch handles all three from a single workflow run.

Connect all ten graded and cropped images to a Collect node that assembles them in the approved spread order. The Collect node feeds three parallel export paths. The Print Export node produces TIFF or PNG files at 300 DPI with an sRGB color profile and a 3mm bleed on all sides — the specification required by most commercial print vendors. The Web PDF node produces a compressed version with a 96 DPI screen resolution, JPEG compression at quality 85, and embedded fonts if the layout includes text overlays. The Social Export node crops each image to 4:5 for Instagram and 1:1 for grid posts, exporting JPEG at quality 92.

Naming conventions matter for handoff. Configure the export node to append the look number and format to each filename: "BrandName-SS26-Look03-Print.tif," "BrandName-SS26-Look03-Web.jpg." This prevents the inevitable confusion when ten images named "image1.jpg" through "image10.jpg" land in a shared drive and no one can tell which is which. Store the export configuration in the workflow so every seasonal refresh uses the same naming pattern automatically.

The following steps describe the complete build sequence for a ten-look lookbook workflow. Each step corresponds to an action in the Floniks editor at /editor. Start with a blank canvas and work from left to right, inputs on the left, exports on the right.

Complete the build in one session and run the workflow on two looks first. Review the output for identity consistency, color grade accuracy, and crop alignment before running the full ten-look batch. Adjust character reference strength or grade parameters as needed, then run the complete workflow and route all outputs to the export branch. The complete run typically completes in under fifteen minutes for a ten-look collection.