How to use SynthID Removal for a supported Google AI image

First, identify which kind of watermark you mean

SynthID is Google’s invisible watermark for AI-generated media. Google describes it as embedded directly into the content and designed to survive common editing operations. It is different from the device, location and date fields that a metadata tool reads.

A visible logo is a third task. Before choosing a tool, establish whether you need ordinary metadata cleanup, the dedicated SynthID image workflow or a visible edit. The wrong workflow can return a valid file while leaving your actual task unresolved.

Source: Google DeepMind: how SynthID works

Check the image before uploading

Spoofzy’s SynthID Removal workflow accepts supported Gemini and Nano Banana images in JPEG, PNG, WebP and HEIC/HEIF formats, up to 20 MB. This Spoofzy workflow is image-only; Google’s wider SynthID technology also applies to other media, which does not mean this tool supports them.

Keep the original. If you have both an original download and a screenshot, use the closest-quality original. A screenshot or repeatedly compressed copy is a different source, and visual detail lost before processing cannot simply be assumed to return.

• Confirm that you created, own or have permission to edit the image.
• Check the extension, actual file type and size.
• Keep the source separate from the result.
• Use Metadata Cleaner only if your task is supported hidden file metadata.

Run the dedicated tool in Telegram

Open the SynthID Removal entry from Spoofzy’s Telegram launcher or the button at the end of this guide. Read the access and AI-credit cost shown before the run. Use that current quote rather than an old screenshot or a copied price from another guide.

Send the image as a file so the workflow starts from the closest source quality available. Follow the bot’s prompts and wait for the returned copy in the conversation. Keep the original until you have completed your review.

Review the returned image with a concrete checklist

Download and open the actual output. Compare it with the source at a useful zoom level and check the subject, fine detail, edges, colour and any text. Confirm that the format and dimensions fit the next step in your workflow.

Visual inspection checks appearance. It does not establish how a particular detector or publishing platform will classify the file. Spoofzy does not promise universal watermark removal, undetectability or any moderation outcome.

• The file opens correctly and is the expected result.
• The subject and important details remain suitable for the intended use.
• Any visible branding or text has been reviewed separately.
• Required attribution and AI disclosures are retained in the publishing workflow.

How to check SynthID after processing

A visual comparison checks appearance. For a watermark check, Google documents uploading one file at a time to the signed-in Gemini app and asking whether Google AI created or edited it. Use files you are permitted to share with Google. Check the original and the downloaded result separately, and save the date and exact responses.

Record the SynthID result separately from any Content Credentials result. A non-detection is not proof of human authorship or a guarantee about another detector. This is a verification method, not a published effectiveness test of Spoofzy; report the observations from your own files without extending them to every image.

Source: Google: verify AI-generated media in Gemini

If the result does not match the task

Check the source type and tool choice before repeating the run. A video, a visible watermark or an unsupported image needs a different path. For a supported file with an unexpected result, contact Spoofzy support with the relevant job context and a description of what differs.

Avoid treating repeated processing as evidence of success. Keep the source and output available for comparison and decide which specific requirement needs to be met before the image is approved.