A Before/After Comparison Workflow

Before/after content is uniquely demanding to produce well. The two frames must share an identical crop and framing so that the comparison is spatially coherent — if the "before" subject is slightly left of center and the "after" subject is slightly right of center, the comparison reads as two different shots rather than a transformation of one subject. Lighting direction and color temperature should also match so that the viewer attributes visible differences to the transformation itself rather than to a change in shooting conditions.

When the transformation is AI-generated — a skincare result, a room renovation, a hair color change, a background replacement — the alignment challenge is built in if the pipeline is designed correctly. The AI transformation node receives the source image and produces an output that matches the source framing because it is conditioned on the source. But when the transformation involves physical changes that alter apparent scale or position (a weight-loss result, a before-installation vs. after-installation product shot, a raw ingredient vs. a finished dish), the two frames will not automatically align and must be corrected in post.

A properly designed Floniks before/after workflow handles all these alignment cases. It stores the source image in an Input node, routes it through whatever transformation is needed, aligns both the source and transformed output to a shared canvas specification, applies the split-line divider and labels, and exports the composite. Every run of the workflow produces the same composite structure regardless of the source subject, which means the before/after format remains consistent across all content in a campaign.

The workflow starts with an Image Input node that accepts the "before" image. This is the source material — the untreated face, the empty room, the raw product, the before-renovation space. The source image should be at the highest available resolution because it will be cropped and composited at the end, and downsampling artifacts in the source half of a comparison are immediately noticeable.

From the Image Input, the source routes along two parallel branches. The first branch holds the source image unchanged and passes it to the Alignment node at the end of the graph — this is the "before" side of the composite. The second branch routes the source through the transformation node or chain of nodes specific to the use case.

For a skincare or beauty result workflow, the transformation is a single AI Image node with a prompt describing the treated result: "same face, same lighting, same framing, skin texture smooth and even, pores minimized, complexion bright and healthy, no heavy filter effect, photorealistic." For a room staging workflow, the transformation chain includes a background replacement or inpainting node followed by a furniture and decor generation node. For a product recolor workflow, the transformation is a recolor node that changes the product color while preserving all other attributes. The transformation output must maintain the same approximate subject position and scale as the source. Include explicit instructions in the transformation prompt to enforce spatial consistency: "maintain subject position, do not reframe, preserve scale and perspective."

After transformation, both the source and transformed images route to an Alignment node. This node takes two inputs and a canvas specification (width, height, and split position), then crops and scales both images to fill their respective halves of the canvas while keeping the primary subject centered within each half. For portrait-orientation comparisons, the canvas is typically 1080x1350 pixels with the split at x=540 — vertical center. For landscape comparisons (room or exterior shots), the canvas is typically 1920x1080 with the split at x=960.

If the source and transformed images have visible positional misalignment — the subject is in a different position in the frame — enable the Alignment node softfocus match option, which uses feature matching to shift the transformed image horizontally and vertically to match the source anchor points before compositing. This handles the majority of drift caused by AI generation and eliminates the need for manual repositioning.

Once both images are aligned to their canvas halves, the Split Compositor node merges them with the divider. The divider is configurable: a simple 3px white line is the most versatile and legible option across content types; a blended transition (a 40px feathered gradient that reveals both images) works well for beauty and skincare comparisons where a hard cut looks clinical; a ragged or hand-drawn divider edge can add editorial character for lifestyle brands. Connect a Label Render node to add "Before" and "After" text in the appropriate quadrants, using the brand typeface and a semi-transparent background pill to ensure legibility on any background color. The composite output from this stage is the master before/after image.

The master before/after composite needs to be delivered in multiple formats for different campaign placements. A Meta paid ad campaign may need 1:1 (1080x1080) for feed and 9:16 (1080x1920) for Stories. An email marketing placement may need 600px wide at 2x resolution for retina displays. A website hero may need 1920x800 cropped from the original 1080x1350 master. Building each of these as a separate manual export creates maintenance overhead; building them as nodes in the export chain means they are produced automatically every run.

Connect the master composite to a Multi-Format Resize node configured with output presets for each campaign placement. Use crop rules that keep the split divider and the most important subject detail visible in each aspect ratio. For the 9:16 crop, shift the canvas upward to show the face or focal object of the comparison prominently. For the wide website hero crop, use a horizontal crop that shows both the before and after halves symmetrically.

For campaigns with multiple before/after pairs — a product launch with several transformation types, or a content calendar with a weekly comparison post — save the workflow as a template with descriptive naming: "BeforeAfter-SkincareCampaign-v1." Each week, open the template, update the Image Input node with the new source photograph, update the transformation prompt if the treated condition differs, and run. The alignment, split composition, labels, and multi-format export all execute automatically. A comparison that might take 30 minutes to produce manually takes under three minutes in the workflow.