Ideogram family
# Ideogram 4.5 Advanced generative model for high-fidelity image creation and precise editing. ## Overview Ideogram 4.5 represents a significant leap in generative AI, specifically engineered for marketing professionals and creative agencies who require both high-quality visual output and granular control over image composition. Unlike standard text-to-image models, Ideogram 4.5 excels at maintaining structural integrity across iterative edits, making it an ideal tool for complex campaign workflows where consistency is paramount. It supports high-resolution generation up to 2K, ensuring that assets are production-ready for digital advertising, social media campaigns, and high-end print collateral. The model's architecture is uniquely optimized for multi-reference guidance. By allowing users to input up to four reference images alongside a text prompt, Ideogram 4.5 enables sophisticated visual synthesis. Whether you are performing interior restyling, modifying product packaging, or adjusting character poses via skeleton references, the model ensures that the untouched regions of your source image remain pristine. This capability significantly reduces the time spent on manual post-production, allowing creative teams to iterate faster and more effectively. ## Use Cases * **Product Photography Variations:** Generate multiple colorways or background settings for a single product while keeping the core product geometry identical. * **Campaign Localization:** Adapt existing ad creative for different regional markets by swapping out background elements or text while preserving the primary subject. * **Interior Design Visualization:** Use reference images of furniture or decor to restyle a room layout without altering the architectural structure of the space. * **Character Pose Consistency:** Utilize skeleton reference images to guide character positioning, ensuring consistent anatomy across different scenes. * **High-Resolution Asset Upscaling:** Leverage the 2K generation capabilities to create crisp, detailed visuals for large-format digital displays. * **Iterative Inpainting:** Precisely modify specific elements within a complex scene using mask-based editing to refine brand assets without regenerating the entire composition. ## Parameters Ideogram 4.5 offers a robust suite of configuration options designed to balance creative freedom with technical precision. Users can adjust parameters ranging from seed-based reproducibility and aspect ratio selection to advanced mask-based inpainting and copyright detection. Please refer to the Input Parameters section in the Pixloop interface to configure your specific generation requirements, including quality settings, output formats, and reference image arrays. ## Tips for Best Results * **Leverage Reference Images:** When editing, use the first reference image as your primary source and subsequent images to guide style or composition for the best alignment. * **Use Precise Masks:** For inpainting, ensure your mask image is high-contrast; white areas define the edit zone, while black areas protect the original content. * **Optimize Output Quality:** For final production assets, set your outputQuality to 95 or higher to minimize compression artifacts. * **Iterate with Seeds:** If you find a composition you like, record the seed value to generate variations that maintain the same layout and structure. * **MagicPrompt for Clarity:** If your initial prompt is sparse, enable the 'magicPrompt' feature to automatically enhance your description with descriptive keywords that align with the model's training data. ## About Ideogram 4.5 is developed and provided by Ideogram, a leader in high-fidelity generative AI. Built upon a proprietary architecture designed for iterative creative workflows, this model is specifically tuned to bridge the gap between abstract text-to-image generation and professional-grade image editing. By focusing on regional integrity and multi-reference guidance, Ideogram continues to set the standard for AI-assisted design in the marketing and advertising sectors.
Model input reference derived from preset schema.
| Name | Type | Required | Default | Description |
|---|---|---|---|---|
| seed | integer | No | null | Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range. |
| prompt | string | Yes | null | Text prompt describing elements to include in the generated output. |
| quality | select | No | "very_low" | Rendering quality. |
| mask_image | image | No | null | Image used to specify which areas of the input image should be edited (UUID, URL, Data URI, or Base64). White marks the area to edit, black is preserved. Applies to the first reference image, must match its dimensions and contain both white and black areas. |
| magicPrompt | select | No | "auto" | Automatic prompt enhancement. |
| aspect_ratio | select | No | "16:9" | Shape of the video. Converted to width and height for the chosen resolution. Ignored when a first-frame image is given (the video follows the image). |
| outputFormat | select | No | "PNG" | Specifies the file format of the generated output. The available values depend on the task type and the specific model's capabilities. |
| numberResults | integer | No | 1 | Number of results to generate. Each result uses a different seed, producing variations of the same parameters. |
| outputQuality | integer | No | 95 | Compression quality of the output. Higher values preserve quality but increase file size. |
| reference_images | array | No | null | List of reference images (UUID, URL, Data URI, or Base64). The first image is the source to edit, the rest guide the edit. |
| copyrightDetection | boolean | No | null | Opt into post-generation copyright detection using Hive likeness and logo checks. |