HiDream
01
Describe the composition
Start with subjects and counts, then placement, background, light and style. Use several short sentences for complex scenes.
Workspace/Image
HiDream-I1
Turn written ideas into images by describing subjects, settings and light.
Estimated costPricing pending
HiDream-I1 · Imagine freely. Create effortlessly.
Your creations
Sign in to see your creations
Describe an idea above to get started.
HiDream-I1 · Capabilities and creative directions
HiDream-I1 is an open-source 17B image generation model using a sparse diffusion Transformer. Its official Full, Dev and Fast variants use 50, 28 and 16 inference steps respectively.
Read the source documentation ↗
02
Explore visual directions
Try photography, illustration or concept design around the same subject. Change style and light while keeping scene details stable.
03
Refine from a result
Once you find a composition, bring the result into E1 to refine its background, materials or colors.
How to get started
Describe your image
Describe subjects, background, composition and light. Bring a result into E1 when you want to edit it.
Choose output settings
Choose the ratio, 1K or 2K resolution and task count, then review the credit estimate before submitting.
Review, download and refine
Open the result to download the original, favorite it or use it as a reference for the next edit.
HiDream-I1 FAQ
Model information, references, credits and managing your work.
What kind of model is HiDream-I1?
The official HiDream-I1 project focuses on text-to-image generation, with 17B parameters and MIT-licensed code and models.
How do Full, Dev and Fast differ?
These are official checkpoint variants with 50, 28 and 16 steps. They describe official inference configurations rather than selectable modes in this composer.
How can I improve object counts and placement?
Specify counts, left/right placement and foreground/background relationships. Simplify unrelated details and address one composition issue at a time.
Which output settings can I choose?
Choose 1K or 2K output and a square, landscape or portrait ratio. Resolution sets the output size tier; the ratio sets the canvas shape.
Can I upload a reference image?
I1 uses text input. Choose E1 to edit an existing image, or O1-Image to create with multiple references.
How are generations charged?
1K costs 6 credits per image and 2K costs 20. Multiply by the task count for the total. Credits are reserved on submission, settled on success and released for failed tasks.
How do I continue editing?
Reuse the prompt to adjust composition. To edit a result, choose Use as reference to open E1 and describe what to change and preserve.
Where can I download and save my work?
Filter Results by image, video or status, open an item to download the original file, and use the star to favorite it. Sign in to the same account to access your history.
What if a result is missing or fails?
Refresh history before resubmitting. Revisit processing tasks later; for failures, check the prompt, references and settings before retrying. Reopen the result if a download is interrupted.
How do I choose output quality and task count?
Start with 1K to review composition and choose 2K when you need more output detail. Batch images are separate tasks; check the total credits before submitting.