Boogu Image Edit: AI Photo Editing by Instruction (Guide)

Boogu Image Edit changes one photo from a plain-text instruction: add or remove an object, change clothes or colours, replace the background, restyle the picture or write English or Chinese text onto it. The result keeps your photo’s framing and proportions; large photos are scaled down to about 1.5 megapixels first. On VdoBloom it runs in the Image Edit tab for 3 credits per edit and takes about 40 seconds.

Disclosure: VdoBloom is our platform.

Most AI photo editors ask you to paint a mask or pick from presets. Boogu Image Edit works the other way round: you upload a single picture and describe the change in a sentence, like “change the grey t-shirt to a red hoodie”. This guide covers what the model is, what it handles well, how to phrase instructions so it changes only what you asked for, what it costs on VdoBloom, and when a different editor is the better pick.

What Boogu Image Edit is

Boogu Image Edit is the editing member of the Boogu-Image-0.1 family, an open-source set of image models released under the Apache-2.0 licence. According to the project site, the work is led by a Huawei research lab in Hong Kong together with several university partners. The family includes a Base model, a fast Turbo model, and two editing models (Edit and Edit-Turbo). All of them are 10-billion-parameter models, and the weights are public on Hugging Face.

The Edit weights were published in June 2026. VdoBloom added the model on September 25, 2026. As of September 2026, the official model card lists these edit types:

• Object insertion, replacement and removal — add a plant to a desk, swap a mug for a glass, remove a bin from a street shot.
• Attribute and material changes — clothing colour, fabric, hair colour, a car’s paint, wood to marble.
• Background and scene replacement — move a portrait from a bedroom to a beach or a plain studio backdrop.
• Style transfer — turn a photo into a watercolour, a comic panel or a clay render.
• Text rendering — write a short line of English or Chinese text into the image, such as a sign, a caption or a label.

It edits exactly one photo at a time. It does not merge two pictures, and it does not create images from text alone.

How to use Boogu Image Edit on VdoBloom

1. Open the Image Edit tab and choose Boogu Image Edit in the model picker.
2. Upload one photo. JPG, PNG and WEBP all work.
3. Type your instruction (at least 2 characters; long, detailed instructions are fine).
4. Press generate. The edit costs 3 credits and the finished image appears in the same panel and in your creations.

A few details that matter in practice:

• Size. There is no aspect-ratio or resolution setting, because the output follows your photo’s proportions. Before the edit, VdoBloom fits the photo within two limits: every side between 512 and 2048 pixels, and no more than about 1.5 megapixels in total. Photos already inside those limits keep their exact size, so a 900 x 1350 portrait stays 900 x 1350. Larger photos are scaled down: a 12 MP phone photo of 4032 x 3024 comes back at about 1414 x 1061, and a 1920 x 1080 frame at about 1633 x 919.
• Speed. In VdoBloom’s integration runs on September 25, 2026, an edit took about 42 seconds. That is slower than several other editors on the site, so plan for it if you are doing a batch.
• Cost. 3 credits per edit, whatever the photo size. New accounts get 10 free credits, which covers three Boogu edits.
• Failures. If the edit fails on the model side, the credits are returned automatically.

How to write edit instructions that work

Instruction editors do best with a request that names one change clearly and says what must stay the same. The model card itself warns that strict preservation of the subject, layout and fine detail is still hard for this model, so spelling out what to keep helps more here than with some older editors.

Say what to change and what to keep

Vague instruction	Clearer instruction
Make his outfit cooler	Change the grey t-shirt to a red hooded sweatshirt. Keep his face, pose, hair, the background and the lighting unchanged.
New background	Replace the background with a quiet beach at sunset. Keep the woman, her clothes and her position in the frame exactly the same.
Add some text	Add the words “OPEN DAILY” in white bold letters on the wooden sign above the door. Do not change anything else.
Clean this up	Remove the plastic bottle from the table on the left. Fill the space with the same table surface.
Make it artsy	Turn the whole photo into a soft watercolour painting with visible paper texture. Keep the composition the same.

More habits that help

• One change per run. “Change the jacket to blue” then, on the result, “replace the background with an office” is more reliable than both in one sentence. Each run is 3 credits, so two clean steps cost 6.
• Point to the thing. Use position and colour: “the white cup on the right”, “the man in the blue shirt”. If there are two similar objects, the model has to guess.
• Put text in quotes and keep it short. Write the exact words in quotation marks and say where they go. The project lists dense typography and small fonts as a known weak spot, so a short headline works far better than a paragraph.
• Describe the result, not the process. “A red hooded sweatshirt” is better than “recolour the pixels of the shirt”.
• Use a clear, well-lit source photo. Small or blurry faces are harder to keep intact. A very small photo is enlarged so its shortest side reaches 512 pixels, which can look soft, so start from the largest original you have.

What it does well and where it falls short

Two edits from VdoBloom’s integration runs on September 25, 2026 show the model’s strong side:

• Clothing swap. A “change the t-shirt to a red hoodie” instruction changed only the garment. The face, pose, background and lighting stayed as they were.
• Background plus text. A request to replace the background and add a line of text kept the person, the framing and the image size intact.

Those are two runs, not a benchmark, but they match the tasks the model was built for: targeted local changes and scene swaps on a single photo.

Honest limits, based on the official model card, the project site and how the model is wired on VdoBloom:

• It is new. Boogu-Image-0.1 is, by its own version number, an early release. Expect some misses and re-runs.
• Identity is not guaranteed. The developers list strict subject preservation, complex body poses and small facial features as open problems. Close-up portraits hold up better than group shots where faces are tiny.
• Dense text struggles. Short signs and captions are fine; long paragraphs or small print tend to break.
• Most stable around 1K. The model card says results are more stable at 1K than at 2K. VdoBloom already scales large photos down to about 1.5 megapixels, so the output is never larger than that.
• It is not fast. About 40 seconds per edit on VdoBloom. If you need many quick iterations, a faster editor may suit you better.
• One photo only. For “put this product into that scene” jobs you need a model that accepts several reference images.

Boogu vs other photo editors on VdoBloom

All of these run in the same Image Edit tab. Prices are VdoBloom credits per image as of September 2026.

How to choose:

• Pick Boogu when you have one photo, want to keep its framing, and the change is a clear local edit. At 3 credits it is one of the cheaper editors on the site. The Boogu Image Edit model page has its specs at a glance.
• Pick Qwen Image 2.1 when you need to paint over an exact area, want a transparent PNG, or have several reference images. Our Qwen Image 2.1 guide covers it in detail.
• Pick GPT Image 2 or Nano Banana Pro for demanding text layouts or 4K results, and accept stricter content filters.

If you are still deciding between editors in general, the AI image editor from photo page lists the other ways to edit a picture on VdoBloom.

Editing photos of real people

Only edit photos of yourself or of people who have agreed to it. Never upload photos of minors for editing, never create intimate or sexual images of real people, and do not use edits of public figures to mislead anyone. VdoBloom checks every request before any credit is charged and blocks sexual content involving minors and non-consensual intimate edits. Boogu carries a “moderate” content label in the model picker: it accepts ordinary edits of adults, such as outfits and backgrounds, but it is not a tool for anything goes. If an ordinary edit is refused, our guide on why AI refuses to edit a photo explains the usual causes.

Frequently asked questions

Is Boogu Image Edit free?

The model weights are free to download under Apache-2.0, but running a 10-billion-parameter model yourself needs a capable GPU. On VdoBloom each edit costs 3 credits, and the 10 free credits on a new account cover three edits.

How long does a Boogu edit take?

About 40 seconds on VdoBloom. An edit of a 900 x 1350 photo took roughly 42 seconds in VdoBloom’s integration runs on September 25, 2026. Larger photos can take a little longer.

Does Boogu Image Edit change my face?

It tries not to, and in a clothing-swap run the face stayed unchanged. It is not guaranteed, though: the developers list small facial features and strict identity preservation as weak points. Add “keep the face unchanged” to your instruction and use a photo where the face is reasonably large and sharp.

Can Boogu add text to a photo?

Yes, in English or Chinese. Put the exact words in quotation marks, say where they go and describe the style (“white bold letters on the sign”). Short phrases work best; small print and long paragraphs are unreliable.

What photo size and file size does it accept?

Upload a normal JPG, PNG or WEBP photo. VdoBloom keeps its proportions and fits it so each side is between 512 and 2048 pixels and the whole image is no more than about 1.5 megapixels. A photo already within those limits keeps its size; a 4032 x 3024 phone photo becomes about 1414 x 1061. The edited image comes back at that size and shape. There is no separate aspect-ratio or resolution option.