An Auto-Crop and Resize Workflow

A social media team managing a brand across six platforms — Instagram feed, Instagram Stories, Twitter, LinkedIn, Facebook, and a web homepage — needs every piece of hero content in six different aspect ratios. The same photograph of a product launch event needs to be a square 1:1 for Instagram feed, a vertical 4:5 for Instagram optimal feed, a 9:16 for Stories, a 16:9 for Twitter and LinkedIn, and a 3:1 panoramic for the website hero banner. That is five manual crop operations per image, and if the photographer delivers 20 images from a shoot, it is 100 individual crop decisions.

The crop decisions are not trivial. A naive center crop for a vertical-to-horizontal conversion will frequently decapitate the subject or cut off a product that is positioned off-center. A good crop requires the editor to look at each image, identify the primary subject, and manually position the crop frame so the subject remains prominent and the composition reads well in the new aspect ratio. At 100 crop decisions, even an experienced editor makes judgment errors that require rework.

An auto-crop and resize workflow in Floniks eliminates this work entirely. The workflow detects the primary subject location in the image, generates a subject-centered crop rule for each target aspect ratio, applies the crops, and for aspect ratios that require a canvas wider than the source image, uses AI outpainting to extend the canvas with coherent content before cropping. The result is a complete, reviewed-quality multi-format export set from a single workflow run.

The first active node in the workflow after Image Input is a Subject Detection node. This node analyzes the source image and returns a subject bounding box — the pixel coordinates of the primary subject (face, product, or focal object) within the image frame. The bounding box is expressed as {subject_x, subject_y, subject_width, subject_height} and is passed as metadata to all downstream crop nodes.

The Smart Crop Rule Generator node receives the source image dimensions and the subject bounding box, then calculates the optimal crop window for each target aspect ratio. The crop rule ensures the subject bounding box is fully contained within the crop window, that the subject is positioned at the rule-of-thirds intersection point closest to the center of the image, and that the crop window does not extend beyond the image boundaries. If a target aspect ratio requires a crop window wider than the source image, the rule generator flags this as an outpainting requirement rather than a crop-only operation.

For portrait photographs, the subject detection model is calibrated to prioritize face detection. For product photography, it prioritizes the product region. For environmental and architectural photography, it falls back to optical center if no dominant foreground subject is detected. You can override the detection priority in the Subject Detection node configuration — set it to "face" for headshot and event photography, "product" for e-commerce shots, or "balanced" for lifestyle photography where multiple subjects share the frame and the composition should be evaluated holistically.

Many multi-format export requirements involve generating a wider crop than the source image supports. A portrait-orientation photograph (4:5 or 9:16) cannot be cropped to a 16:9 web hero without either severely cropping the subject or extending the canvas into territory the camera never captured. For these cases, the workflow routes the image through an Outpainting node before applying the wide-format crop.

The Outpainting node receives the source image and the required canvas extension — the number of pixels to extend on the left side, right side, or both — from the Smart Crop Rule Generator. It synthesizes new image content that fills the extended canvas in a style visually consistent with the source: "continue the background environment of the existing image into the extended canvas area, match the lighting direction and color temperature, maintain the same depth of field and background blur, produce seamless edges with no visible boundary." The quality of outpainting is high enough for background extension (sky, interior walls, studio backgrounds, outdoor environments) but should be reviewed when the extension area might include complex foreground content.

After outpainting, the extended canvas passes to the standard crop node, which applies the 16:9 or panoramic crop rule using the subject position from the earlier detection pass. The result is a properly framed wide-format image where the subject is correctly positioned and the extended background areas are visually coherent. For campaigns where outpainting quality is critical (premium brand imagery, luxury product photography), the workflow can be configured to flag outpainted outputs for human review before they join the export batch.

The final section of the workflow is the export chain. A Platform Export node accepts the cropped outputs from each aspect ratio branch and applies platform-specific encoding settings: JPEG quality 92 for web and social (preserving margin above platform recompression), PNG for any context where text overlay sharpness matters, and WebP for modern web deployment contexts where file size is a priority. The node also applies the correct pixel dimensions: 1080x1080 for Instagram square, 1080x1350 for Instagram portrait, 1080x1920 for Stories, 1920x1080 for LinkedIn and YouTube cover, and custom dimensions for email and web hero placements.

File naming is handled by a Naming Template node that receives the original filename and the platform label and produces an output filename in a consistent format: "{original_name}{platform}{dimensions}.jpg." For batch workflows processing a full shoot of 20 or more images, the naming template ensures files are organized correctly when they land in the export folder and can be sorted by platform without additional renaming work.

Save the completed auto-crop and resize workflow as a Floniks template. The template needs no modification between uses — just drop in a new source image via the Image Input node and run. The subject detection, crop rule generation, outpainting where required, and multi-format export all execute automatically. For teams processing a full week of social content on Monday, running all source images through the workflow takes less time than manually cropping a single image, and the output quality is consistent across every format.