Creative · Community report
Remove an object from a photo
One person on X shows Muse calling SAM 3.1 through a Meta Model API key to cut a dresser out of a photo, then using Muse Image to fill the gap. Only that one example is shown, so below: how to connect the models and check each edit before you remove an object from a photo of your own.
What you can get
What it is
Edited photo
A photo with the chosen object removed and the background filled in, for you to review.
It came back right when
- The named object is gone from the result.
- The filled-in area doesn't show obvious artifacts, warping or mismatched lighting.
- The rest of the photo is unchanged.
- The result was shown to you, not posted or shared automatically.
- You decided, separately, whether an edited photo is appropriate for wherever it's headed.
The checks are editorial, not from the source.
- Confirm the connectionStops for your approval
A connected image-model account is set up and working.
- Name the object
You point to the specific object to remove.
- Segment and fill
Segments the object and fills the background.
- Review the resultStops for your approval
You check the fill before using the photo anywhere.
Editorial, not from the source
Works when
- You have a photo with one clear, identifiable object to remove.
- The background around the object is simple enough to plausibly fill in.
- You just need the result for personal use, not an official document.
Does not work when
- The object overlaps a face, text, or something official-looking.
- You need the edit to be forensically accurate, not just visually plausible.
- You haven't set up a connected image-model account.
Remove an object from a photo: review template
| Area | What changed | Your check |
|---|---|---|
| [Removed object] | [Segmented and cut out] | [Confirmed gone] |
| [Fill area] | [Background reconstructed] | [No visible artifacts] |
| [Rest of photo] | [Should be unchanged] | [Matches original] |
The source case
On X, @nikhilaravi describes connecting Muse to a Meta Model API account by pasting an API key into Muse's connector vault, giving it access to models including SAM 3.1 and Muse Image. In the example shown, Muse called SAM 3.1 to segment a dresser out of a photo, then used Muse Image to fill in the missing background where it had been. One example is shown; checked on 2026-10-09, not a report of how the result holds up across different photos or objects.
A public report by someone outside this library. Not independently reproduced here.
From the source
Source excerpt
Here I wanted to cut this dresser out of a photo. Muse called SAM 3.1 to segment it, and then I asked it to use Muse Image to complete the missing regions on the wooden top where the objects had been.
Quoted from the original posts. Not written by this guide, and not a result produced here.
Suggested prompt
Bring The photo, Which object to remove. Acts Edits a copy of the photo you provide. Does not post or share the result unless you ask. Source community report, checked 2026-10-09.
Written by this guide as a starting point, not quoted from the source. Your edits stay in this browser.
Other ways to ask
Editorial rewrites of the brief above, for a different emphasis. Copy them separately.
What you need
Editorial, not from the source
You supply
- The photo
- Which object to remove
Connected
- A connected image-model account, such as a Meta Model API key in the connector vault
Approved by you
- Your review of the edited photo before it's used anywhere
How to try it
Editorial, not from the source
- Connect the image-model account it will use, the post describes a Meta Model API key pasted into the connector vault, before asking for anything.
- Point to the specific object in the photo you want removed, rather than a vague area, so segmentation has a clear target.
- Ask for the result shown to you, not saved or shared anywhere, before you decide whether the fill looks right.
- If the fill looks wrong, say which part to redo rather than starting over on the whole photo.
Set up the connection before you ask for anything
This task doesn't come from a built-in eraser; it depends on a connected image model. The source describes pasting an API key for a Meta Model API account into Muse's own connector vault, after which Muse can call specific models, SAM 3.1 for segmentation and Muse Image for filling in a background, on request. Nothing here happens without that connection existing first, so the real first step is confirming the account is set up and linked, not asking for the edit and hoping a capability appears. If you don't have a model account connected, the honest answer is that this specific workflow isn't available to you yet, rather than that Muse can't do photo editing at all; it depends on which models you've connected and what they're each built to do.
Point to the object, not the area
Segmentation works by identifying a specific thing in an image, so the clearer you are about what that thing is, the better the model has to work with. In the example behind this page, the object was a dresser against a fairly plain background, a reasonably easy case: one identifiable piece of furniture, simple surroundings. Say exactly what you want removed rather than describing a region of the photo in vague terms like 'that part on the left,' since an ambiguous instruction gives the segmentation step more room to pick the wrong thing, or part of the wrong thing, which then carries through to a fill that's solving the wrong problem from the start.
Judge the fill before you trust it
Removing an object is really two steps, cutting it out and then filling the gap convincingly, and the second step is where results vary the most. A plain wall or floor fills in easily; a patterned surface, a reflection, or a shadow the object was casting is much harder to reconstruct believably, and none of that is covered by the one example this page is built on. Look closely at the filled area against the rest of the photo before you use the result anywhere, for warping, mismatched texture, or lighting that doesn't line up, and if you're planning to use the photo for something where accuracy matters, an insurance claim, a listing, identification, decide separately whether an edited photo is even appropriate for that purpose.
Open Muse to try it. That is the product's own site; this guide is not part of it.
Limits and confirmation points
Limits to confirm
- One example, one object type, furniture against a plain surface. Faces, busy backgrounds, reflections and shadows aren't covered and may fill in less convincingly.
- This depends on a connected third-party model account; without one set up, the capability described here isn't available.
- The post describes pasting an API key into a vault 'that Muse never sees,' but doesn't describe what happens if the connection fails or the key is revoked.
- Removing an object can also remove context a photo needs to be trusted, for identification, insurance or listings; check whether that matters before using an edited photo for anything official.
Stop and check
Editorial, not from the source
- Before it starts: confirm the image-model account is connected and working.
- Before you use the result: check the filled-in area against the rest of the photo.
- If you plan to share it: decide whether an edited photo is appropriate for that use.
If it goes wrong
Editorial, not from the source
- The fill looks obviously wrong where the object was
- Point to the specific area and ask for that region to be redone rather than the whole photo; give more context about what should be there.
- The connector won't link to the image-model account
- Re-check the API key and vault connection described in the source; a broken connection means the capability isn't available until it's fixed.
- It removes or changes something you didn't ask about
- Be explicit that only the named object should change, and ask for the rest of the photo to be left untouched.
Common questions
Does this work without a connected model account?
No. The source describes connecting a Meta Model API account via an API key in Muse's connector vault; without a connection like that, this specific workflow isn't available.
What kinds of objects does this work best on?
The one example shown is furniture against a fairly plain background. Faces, patterned surfaces, reflections and shadows are harder cases not covered by that example.
Will it change anything else in the photo?
It shouldn't, but check. Ask explicitly for only the named object to be removed and the rest of the photo left untouched, then compare the result to the original.
Can I use the edited photo for something official, like a claim or listing?
That's a separate decision this directory can't make for you. Removing an object can remove context a photo needs to be trusted for identification, insurance or listing purposes.
How do I fix a bad fill?
Point to the specific area that looks wrong and ask for just that region to be redone, with more description of what should actually be there, rather than restarting the whole edit.
Next
Related to this task
More in Creative
Similar tasks
Get recurring task recommendations — ask your own agent to check for relevant updates.