​

DreamCraft: Explore AI Imagery with HiDream

​

Overview

DreamCraft is a versatile AI image tool built upon the open-source HiDream-I1 model and its editing counterpart, HiDream-E1. It offers two primary modes: Generate, for creating images from text prompts, and Edit, for performing maskless modifications on existing images. DreamCraft provides access to different “flavors” of the HiDream-I1 model (Fast, Balanced, Quality) and leverages the E1 model for its editing capabilities, making advanced AI image creation and modification more accessible.

​

Modes of Operation

DreamCraft offers two distinct modes to suit your creative workflow:

​

Generate Mode: Create Images from Text

In Generate mode, you can create entirely new images by providing a text prompt. You can select from different versions (“flavors”) of the HiDream-I1 model to prioritize speed or quality.

Inputs & Options (Generate Mode):

• Prompt: Your text description of the desired image.
• Negative Prompt: Specify elements to omit.
• Image Size: Select the resolution and aspect ratio.
  • Possible values: square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9, custom (with width/height inputs).
• Num Inference Steps: Controls detail (1-50, default 25). Note: Max steps for Fast model is 12.
• Seed: For reproducible results (optional).
• Guidance Scale: Adjusts adherence to prompt (0-20, default 5). Note: Only available for Quality model.
• Number of Images: Number of variations to generate (1-4).
• Model: Select the HiDream-I1 model version.
  • Possible values: Fast, Balanced, Quality.
• Output Format: File format (JPEG, PNG).

​

Generate Mode Examples

Explore the diverse range of images you can generate from text prompts:

A paper quilling, layered paper scene of an astronaut playing an electric guitar while floating in space, surrounded by stars and planets, capturing the cosmic energy and musical vibe in zero gravity.

Parrish, Yorkshire canal, canal boat docked at water’s edge, gently rippling water, canal banks dotted with lush trees, traditional Yorkshire industrial buildings, weathered brick facades, vintage chimneys, rustic charm, nostalgic, cinematic lighting, ultra-detailed, hyper-realistic, serene atmosphere, early morning mist, reflective water, high resolution

The image depicts a beautifully ornate, oval-shaped clear glass sphere with an intricately designed frame. The frame is made of a golden material with elaborate, swirling patterns. Inside the sphere there is a detailed, red and white dragon emerging from what appears to be water or waves made of crystal. The dragon has a majestic and mythical appearance, with flowing, spiky hair and a fierce expression. The background of the sphere shows a dynamic scene with splashing water and waves, adding to the sense of movement and energy. The sphere is mounted on a tree trunk, and there are pink flowers and green foliage surrounding it, enhancing the natural and mystical atmosphere of the scene.

masterpiece of a photographer Extreme close-up of tiny miniature workers in colorful construction uniforms and hard hats industriously painting oversized Easter eggs, workers using tiny buckets of vibrant paint, miniature ladders, ropes and pulleys to access different parts of the giant eggs, some workers carefully painting intricate patterns while others roll large eggs through puddles of dye, detailed diorama-like scene with perfect depth of field highlighting the meticulous work process, some workers carrying paint brushes bigger than themselves, droplets of paint glistening under studio lighting, vibrant color palette with diverse egg designs being created, photorealistic macro photography with tilt-shift effect, highly detailed miniature construction site with tiny scaffolding, microscopic paint rollers and diminutive paint splashes, whimsical Easter-themed miniature world captured with 8K resolution and crystal clarity.

​

Edit Mode: Maskless Image Modification

Edit mode allows you to modify an existing image simply by providing written instructions. This uses the HiDream-E1 model, which is designed for image-to-image editing without requiring manual masking.

Inputs & Options (Edit Mode):

• Edit Instruction: Your text description of the desired changes to the image.
• Target Image Description: Describe what the final image should look like (optional).
• Edit Image: Upload the base image you want to edit.
• Negative Prompt: Specify elements to avoid in the edited image.
• Edit Num Inference Steps: Controls detail of the edit (1-50, default 50).
• Edit Guidance Scale: Adjusts adherence to edit instruction (0-20, default 3.5).
• Image Guidance Scale: Controls how much the original image influences the edit (0-10, default 2).
• Edit Seed: For reproducible results (optional).
• Edit Num Images: Number of edited variations (1-4).
• Output Format: File format (JPEG, PNG).

​

Edit Mode Example

See how DreamCraft can modify an existing image using written instructions:

Edit Instruction: add realistic dark clouds in the sky
Target Image Description: mountains
Negative Prompt: low resolution, blur

​

Understanding the HiDream Model

DreamCraft is powered by the open-source HiDream model series, which includes the HiDream-I1 (for generation) and HiDream-E1 (for editing) models.

​

Mastering Prompts for DreamCraft

Prompting in DreamCraft varies based on the mode.

​

Prompting in Generate Mode:

​

Prompting in Edit Mode:

​

How to Use DreamCraft

Embark on your creative journey with this general workflow:

​

Input Parameters and Options

DreamCraft offers a range of input parameters, varying based on the selected Mode.

​

Generate Mode Parameters:

​

Edit Mode Parameters:

​

Tips for Best Results

​

Conclusion

DreamCraft provides a powerful and accessible platform for AI image creation and editing through the open-source HiDream model. With its distinct modes and intuitive controls, it empowers users to generate new visuals and modify existing ones with remarkable flexibility.