Re-crops an image to a target aspect ratio without inventing new pixels. Use when the subject already fits and only the framing changes.
# Image Reframer Intelligently adjust aspect ratios while preserving composition. ## Overview The Image Reframer model is a specialized AI-driven tool designed to solve the common marketing challenge of adapting visual assets for multiple platforms. Whether you need to convert a landscape hero image into a vertical social media story or a square product shot into a cinematic banner, this model intelligently recalculates the image composition. Unlike simple cropping, which often loses critical subject matter, Image Reframer analyzes the original image to maintain the subject's position, pose, and perspective, ensuring the visual narrative remains intact across different aspect ratios. This model is distinct because it balances structural integrity with creative flexibility. By leveraging advanced diffusion-based image editing, it fills in the gaps created by aspect ratio changes, allowing for seamless expansion or reframing. It is an essential utility for ad agencies and content creators who need to repurpose high-quality photography for cross-channel campaigns without manual editing or complex Photoshop masking. ## Use Cases * **Social Media Adaptation:** Instantly convert horizontal website banners into vertical 9:16 formats for Instagram Reels and TikTok ads. * **E-commerce Optimization:** Reframe product photography to fit standardized 1:1 or 4:3 grid layouts for catalog consistency. * **Display Advertising:** Quickly generate multiple ad sizes (e.g., 21:9, 3:2) from a single master creative to support programmatic display campaigns. * **Editorial Layouts:** Adjust hero images for blog posts or landing pages to ensure they fit specific design templates without cutting off key visual elements. * **Brand Asset Repurposing:** Transform legacy marketing assets into modern aspect ratios to maintain brand visual standards across new digital touchpoints. ## Parameters This model offers granular control over the reframing process. Users can adjust the seed for reproducibility, define the target aspect ratio, and manage technical output settings like format and inference steps. Please refer to the input parameter schema for specific data types and default values. ## Tips for Best Results * **Start with High-Resolution Source Images:** The model performs best when the input image has clear subject definition and high detail, as it uses this information to guide the reframing process. * **Experiment with Guidance Scale:** If the reframed output feels too detached from the original, lower the guidance scale; if you want more creative freedom in the newly generated areas, increase it slightly. * **Use Consistent Seeds:** If you are testing different aspect ratios for the same campaign, lock your seed to ensure the model maintains a consistent aesthetic across all generated variations. * **Mind the Safety Tolerance:** If you are working with sensitive brand imagery, adjust the safety tolerance to ensure the output aligns with your organization's content policies. * **Iterate with Inference Steps:** For complex compositions, increasing the number of inference steps can help the model better blend the original content with the newly generated frame extensions. ## About The Image Reframer is powered by the fal-ai/image-editing/reframe model. Developed by the team at fal, this model represents the cutting edge of image-to-image editing technology, designed specifically for high-throughput creative workflows. It is part of a broader suite of generative AI tools that prioritize speed, quality, and commercial-grade reliability for marketing teams and creative professionals.
Model input reference derived from preset schema.
| Name | Type | Required | Default | Description |
|---|---|---|---|---|
| seed | number | No | null | The same seed and the same prompt given to the same version of the model will output the same image every time. |
| image_url | image | Yes | "" | URL of the old or damaged photo to restore. |
| sync_mode | boolean | No | false | If set to true, the function will wait for the image to be generated and uploaded before returning the response. This will increase the latency of the function but it allows you to get the image directly in the response without going through the CDN. |
| aspect_ratio | select | No | "16:9" | The desired aspect ratio for the reframed image. |
| output_format | select | No | "jpeg" | The format of the generated image. |
| guidance_scale | number | No | 3.5 | The CFG (Classifier Free Guidance) scale is a measure of how close you want the model to stick to your prompt when looking for a related image to show you. |
| safety_tolerance | select | No | "2" | The safety tolerance level for the generated image. 1 being the most strict and 6 being the most permissive. |
| num_inference_steps | number | No | 30 | Number of inference steps for sampling. |