Hunyuan 3.0 Image to Image: Edit, Restyle and Merge Photos

Last verified: October 19, 2026

Hunyuan 3.0 edits a picture you upload so that it follows a written instruction. If you add a second picture, it can take elements from both and combine them in one new image. That makes Hunyuan 3.0 image-to-image editing a good fit for changes you can describe in one sentence: a new setting behind a product, a street photo redone as a print, a vase added to a shelf, or a stray object removed from a scene.

The model looks at the input image and works through your instruction before it generates anything. A request like “make it a winter evening” is applied to what is actually in the frame instead of being treated as a description of a brand-new picture. This matters most when the result has to stay true to a real product, room or location. You are changing an existing picture, not hoping a text description will recreate it.

Most of the work comes from the picture and the prompt. Frame shape and file type are set just before you generate, and an optional seed field sits next to them. The steps below go through each control, and the writing tips explain how to word an instruction so the model changes exactly what you mean.

Explore more image to image

Image To Image Seededit V3 Nano Banana 2 GPT Image 2 Seedream 4.5 Edit Qwen Image 3.0 Qwen Image 3.0 Pro
Capability Snapshot

Hunyuan 3.0 Inputs and Outputs at a Glance

What the form takes in and what it gives back on each run.

Input

Up to 2 images (JPG, JPEG, PNG or WEBP, up to 10 MB each) + a text prompt of up to 2,048 characters

Aspect Ratio

1:1, 3:4, 4:3, 9:16, 16:9

Output Format

JPEG or PNG

Output

1 image per run

How to Write Edit Instructions for Hunyuan 3.0

Name One Action and One Target per Edit

Build each instruction around one verb and one clearly named object: “remove the parked scooter on the left”, “add a ceramic vase of tulips to the side table”, “change the jacket to burgundy corduroy”. Adding, removing and changing elements are the edits the model is built for, and areas outside the target are meant to stay intact. The more precisely your prompt names the target, the more of the original frame you keep.

State the Goal of the Edit, Then Add Specifics

Tell the model what the picture is for, not just what to change. “Make it a holiday gift-guide shot” gives it more to work with than “add decorations”. The model uses broad world knowledge to work out what you mean and fills short requests with details that suit the context. If the first result is too generic, name the exact props, colors or materials you want and run it again.

How to Edit an Image with Hunyuan 3.0

Five steps take you from upload to a downloaded edit.

Editing a Photo vs. Starting from Text with Hunyuan 3.0

Which Hunyuan 3.0 workflow fits depends on whether you already have a picture to start from.

What you start from This form starts from a picture you upload plus a written instruction describing the change. The Text To Image workflow builds the picture from a written prompt. Use this form when a real photo, product shot or sketch has to shape the result, and Text To Image when the idea exists only in words.
Keeping a real subject recognizable Edits are designed to keep the key visual elements of the uploaded picture while the instruction changes the rest. Text To Image creates every element from the prompt's description, so the subject is whatever the words define. Choose this form for an existing product, room or place that has to look like itself, and Text To Image for new concepts that have no reference picture.
Stay on this form whenever an existing picture should shape the result; switch to Text To Image when you are starting from a blank canvas.

Common Hunyuan 3.0 Editing Problems and Fixes

What usually goes wrong with an edit, and what to change before you run it again.

My photo won't upload in Add Image

Add Image checks each file's format, file size and minimum pixel size, and rejects anything outside the limits. The upload question in the FAQ below lists the exact limits. The usual problems are HEIC exports from phones, very large camera files, and small thumbnails or cropped screenshots. Export the picture again in an accepted format at a moderate size, then add it and carry on. Nothing else in the form needs to change.

The Prompt box won't let me type any more

The Prompt box has a character limit. Long instructions often hit it because they describe the whole picture all over again. Cut sentences about things already visible in your upload. Keep the words that describe the change, its target and anything that must stay as it is. Then generate as usual.

With two images, it took the wrong element from the wrong picture

If your prompt doesn't say what each upload contributes, nothing tells the model which picture plays which role. Refer to the pictures by their order and give each one a job. Tencent's own two-image example does exactly this: it uses the logo from the first image and the fridge-magnet material from the second. Rewrite your instruction the same way, for example “put the jacket from the first image on the person in the second”, and run it again.

I only get one image and want options to choose from

Each run returns 1 image, so you get alternatives by running again. Keep the prompt the same, type a different number in Seed for each new run, and compare the results side by side. Once one direction looks right, refine it by adjusting the wording.

Frequently Asked Questions

Which image files can I upload to Hunyuan 3.0, and how many?

Add Image accepts JPG, JPEG, PNG and WEBP files of up to 10 MB each, at least 300×300 pixels in size. Each run needs at least one image and accepts up to two. The second is for edits that combine elements from both pictures.

Can I use Hunyuan 3.0 without uploading an image?

Not with this form. Add Image is required, because every edit starts from a picture you provide. To generate from a written description alone, use the Hunyuan 3.0 Text To Image workflow instead.

Can I keep editing the image Hunyuan 3.0 returns?

Yes. Download the result, add it in Add Image as the starting picture for your next run, and describe only the next change. Building an edit in stages keeps each instruction short, and you keep every in-between version in case a later step goes wrong.

What makes one Hunyuan 3.0 generation cost more than another?

The credit cost of a run is calculated from the options you choose and shown above the Generate button once you are signed in. Aspect Ratio is the setting that changes it, so check the figure again after you switch ratios. Asking Enhance Prompt for suggestions doesn't use credits.

References

Sources and citations used to support the content provided above.

Updated: 2026-10-19 23:38:03 3 Sources

x.com

Source Link
https://x.com/TencentHunyuan/status/2015635861833167074

github.com

Source Link
https://github.com/Tencent-Hunyuan/HunyuanImage-3.0

huggingface.co

Source Link
https://huggingface.co/tencent/HunyuanImage-3.0