Upload one or more reference images, then use a prompt to change style, setting, light, and detail while keeping the subject and composition recognizable.
From precise edits to style transformations, choose the model that fits your creative vision.
No Images Generated
Load a direction in one click
Pick a visual direction to prefill the generator. You can edit every detail before uploading your own reference image.
Keep identity and add a light, hand-painted finish.
Rebuild a portrait with clean professional lighting.
Keep the person and wardrobe, elevate the art direction.
Lock product details and rebuild the commercial scene.
Turn an everyday frame into a narrative keyframe.
Keep the subject and introduce snow and cool light.
Turn the subject into a complete animated character.
Build a consistent set around the product and brand color.
Five practical transformations
Choose a direction to load an editable prompt. Add your own image and the model will generate from both the reference and your instructions.
Cinematic portrait
Keep identity, replace setting and light
Product scene refresh
Keep the product, rebuild material and lighting
Brand visual system
Carry brand cues into a new composition
Keyframe rework
Anchor the subject and explore a new film world
Studio portrait upgrade
Keep identity and create a polished portrait
Four core image-to-image workflows
From templates and subject preservation to multiple references and visual variants, every path returns to the same working generator.
01 / TEMPLATEChoose a visual template close to your goal, then replace the example with your own reference. The prompt remains fully editable, so you do not have to guess from a blank field.
02 / PRESERVEName the features that must not change—such as a face, package, outfit, or pose—then describe the new setting, camera, and lighting.
03 / MULTI-REFERENCEProvide separate references for a person, wardrobe, or mood, then explain the role of each image in the prompt. Current models support 1 to 3 references.
04 / VARIANTSUse a successful result as the next reference, then vary setting, color, and camera over several rounds. This creates a controlled series without pretending one click produces a full batch.
Keep exploring image workflows
These entry points reuse yeha.ai's real generation and editing capabilities while narrowing the interface to one specific outcome.
Create a new image from a written idea.
Open tool
Develop a keyframe into a moving shot.
Open tool
Rebuild product setting, light, and commercial finish.
Open tool
Remove a background while keeping a clean subject edge.
Open tool
Improve clarity and detail in a low-resolution image.
Open toolConvert a personal photo into a distinct avatar style.
Open toolTrust and privacy
You can see the selected model, reference limit, and estimated credits before generation. See our Privacy Policy and Terms for handling of media and account data.
Only media you select is added to the current generation task, and the upload count follows the selected model's real capability.
Each yeha.ai model name is paired with its underlying model instead of hiding the generation path behind a vague label.
Model cards and the generate button show estimated credits, while tasks and results continue to use your existing history.
Compact pricing
This section shows current monthly plan prices and credits. Full benefits, annual discounts, and payment options are listed on the pricing page.
FAQ
Real generation, subject preservation, multiple references, models, and continued editing.
Yes. The first section reuses yeha.ai's live image generator, including real uploads, model selection, authentication, credits, task polling, error states, downloads, and continued editing.
Text-to-image creates from words alone. Image-to-image combines your prompt with an uploaded reference that can guide subject, composition, or style.
Models usually respond to subject, silhouette, composition, color, and style, but they do not guarantee pixel-perfect duplication. State the identity, package details, or pose that must remain.
Common tasks include moving a person to a new setting, refreshing product photos, restyling, unifying brand visuals, exploring film keyframes, and extending asset sets.
It depends on the model. Current selectable models support 1 to 3 references, and the uploader enforces the selected model's actual limit.
Use a clear reference with a complete subject, then describe both what must stay and what should change. Changing fewer variables per round is usually more stable.
Try Yeha Pro for natural-language and multi-reference work, Yeha Ultra for photoreal commercial imagery, Nano Banana for fast general edits, and Nano Banana 2 for complex instructions and stronger structure.
Yes. You can download a result, use it as the reference for another image-to-image round, or continue in the image creation workspace.
Start from a reference
State what must stay, then describe the new style, background, and light. Pick the model that fits the task and keep iterating from the result.