Free AI Tool · Architects & Designers
Smart Crop for Renders — AI Reframing for Social, Print & Presentations
Got the wrong shape from ChatGPT, Nano Banana or your render engine? Drop the image here. The AI finds the subject and proposes six crops at exact ratios — for Instagram, LinkedIn, slides and print sheets — then your browser cuts them from the original at full resolution.
By uploading an image you agree to our Terms of Service and Privacy Policy.
NSFW or sensitive content is strictly prohibited.
Fix the aspect ratio of a ChatGPT or Nano Banana image
Generated an image in ChatGPT or Gemini, asked for 16:9 or 4:5, and got a near-square anyway? That is how those tools work: they generate at a handful of fixed sizes, and the ratio in your prompt is a request, not a setting. They are very good at generating — they just do not give you ratio control.
This tool finishes the job. Upload the image, and the AI locates the subject — the building, the facade, the focal point of the room — and proposes six crops at exact ratios that keep it in frame. The crop is cut from your original pixels, so nothing is regenerated and nothing loses detail. It works on any image regardless of where it came from: ChatGPT, Gemini, Nano Banana, Midjourney, a render engine, or a photo.
Want to understand the why — which ratios each model supports and what pixel size they produce? The aspect ratio and print format guide has the full tables.
How it works
Three steps. No account needed.
Upload the image
Any render, AI image or photo — JPG, PNG, WebP or HEIC. A small copy goes to the AI for analysis; the full-resolution file stays in your browser.
Say what it is for
Print (landscape, portrait or both), social media (Instagram, LinkedIn or both), or a mix of everything. The AI finds the subject and proposes six different framings for that purpose, each with a one-line reason.
Download at full resolution
Each crop is cut from your original file on a canvas in your browser and downloaded as a JPG. The real output pixel size is shown on every preview.
How the crop is chosen
The AI never draws the rectangle. It returns an anchor point — where the subject is — plus a target ratio and a destination. Your browser then computes the largest rectangle at exactly that ratio, centred on the anchor and pushed back inside the image if it overflows.
That split is deliberate. Vision models are good at "roughly where is the subject" and unreliable at precise bounding boxes — they return rectangles that spill out of the frame, at the wrong ratio, or collapsed to a sliver. Keeping the geometry in code means the crop is always at the exact ratio and always inside the image; a mediocre anchor produces a slightly off-centre crop, never a broken one.
The ratios on offer are the eight AI render models output natively: 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16 and 21:9. If a crop comes out below 800 px on its short side, it is flagged as social-only, with a link to upscale the render first.
Crop reference by destination
Where each crop goes, the pixel size the platform expects, and the small trims that a few destinations still need because no model outputs 4:5, 1.91:1 or 1:√2 directly.
| Destination | Ratio | Pixels | Recipe |
|---|---|---|---|
| Instagram square post | 1:1 | 1080×1080 | Native — crop straight to 1:1 |
| Instagram portrait post | 4:5 | 1080×1350 | Crop to 3:4, then trim 6.25% off the height in Instagram |
| Instagram Story / Reel, TikTok, Shorts | 9:16 | 1080×1920 | Native — crop straight to 9:16 |
| LinkedIn / Facebook link preview | 1.91:1 | 1200×628 | Crop to 16:9, then trim 6.9% off the height |
| LinkedIn banner | 4:1 | 1584×396 | Crop to 21:9, then trim the height to 4:1 — a big cut, pick a wide subject |
| Presentation slide, YouTube, website hero | 16:9 | 1920×1080 | Native — crop straight to 16:9 |
| ISO A sheet — A4, A3, A2, A1, A0 | 1:√2 | see guide | Crop to 3:2 (or 2:3), then trim 5.7% off the long side when placing on the sheet |
| ARCH B / ARCH D sheet | 2:3 | see guide | Native — crop straight to 2:3 or 3:2 |
For the pixel sizes of every sheet format at 150 and 300 DPI, see the aspect ratio and print format guide.
Frequently Asked Questions
About the Smart Crop tool
- Yes, and it is the main reason this tool exists. Those models generate at a small set of fixed sizes, so the ratio you asked for in the prompt was a request rather than a setting — you got the nearest size the model had. Upload the image here and the AI proposes six crops at exact ratios, keeping the subject in frame. The tool works on any image regardless of where it came from: ChatGPT, Gemini, Nano Banana, Midjourney, a render engine, or a photo.
- The vision model looks at the image, names the primary subject — the building, the facade, the focal point of the room — and returns an anchor point for it plus a suggested target ratio and destination. Your browser then computes the crop rectangle around that anchor: the largest rectangle at exactly the requested ratio that stays inside the image. That split matters. The model handles what it is good at, roughly locating the subject, and the geometry is arithmetic, so the crop is always at the exact ratio and always inside the frame.
- It removes pixels, it does not degrade the ones that remain. A crop takes a region of the original at full resolution, so the detail is identical — the file is simply smaller. What matters is whether what is left is big enough for where it is going: 1080 px on the short side is plenty for social, but a print at 300 DPI needs roughly 1240 px for an A4. Each crop shows its real output pixel size, so you can see exactly what you get. When a crop falls below a useful print size the tool says so and links to the upscaler.
- Yes. Each run proposes six crops for six different destinations, and you download each one separately at full resolution, so a single render can leave with a 9:16 Reel, a 1:1 feed post and a 3:2 crop for a printed sheet in one go. Note that an A3 sheet is 1:√2, which no model outputs natively — the closest crop is 3:2, and you trim about 5.7% off the long side when placing it on the sheet.
- The eight ratios RenderAI models render natively: 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16 and 21:9. Sticking to that set keeps the crop consistent with anything you generate afterwards. For destinations outside it — 4:5 for an Instagram portrait post, 1.91:1 for a link preview, 4:1 for a LinkedIn banner — crop to the closest native ratio here and trim the small remainder in the platform's own editor.
- JPG, PNG, WebP and HEIC. HEIC files from an iPhone are converted to JPG automatically before anything else happens. The tool works on any kind of image — a render, an AI-generated image, a photo of a model, a hand sketch — because the crop is computed from the pixels you upload, not regenerated. Downloads always come out as JPG at the crop's full pixel size.
Other Free Tools
From ratios and print sizes to scales and units — these tools cover the practical workflows architects and interior designers deal with every day.
Liked the crops?
Imagine what RenderAI does with the whole render
A crop reframes what you already have. RenderAI creates the image itself — from a sketch, a plan or a reference — at the exact ratio you need, so there is nothing left to crop.
Try RenderAI