Troubleshooting
Why was my generation declined?
Sometimes a generation stops before it starts, or the model refuses it. A declined generation is never charged: the credits held for it go straight back to your balance.
Updated ·
Two kinds of check
Ours, before anything is sent
If a picture shows a real-looking person, we ask you to confirm you have the right to use their face and likeness, or that the person was made by AI. A well-known public figure, or someone who looks under 18, can't be used from a picture, and a prompt that asks for a real public figure is refused the same way. Nothing has been sent to a model yet.
The model's, after it is sent
Every model's maker runs its own content policy on the prompt, the pictures and sometimes the finished result. When it says no, we tell you the model declined. The maker doesn't tell anyone which word or detail set it off, and no one can switch its policy off.
What the messages mean
“The model declined this prompt or image under its content policy”
The model's filter read something in the prompt or a picture as sensitive. Change what it may have misread (below) and try again.
“This model refused the picture as a possible real person”
The model takes a realistic face in a picture for a real person's photo. AI-made people need portrait approval. Real people need verified, authorised likenesses registered for this workspace. Approval remains subject to review.
“A picture in this request shows a real person”
Our own check. Confirm you have the right to use that person's likeness, or that the person was made by AI and is nobody real, and send it again. Your confirmation is kept with the result.
“This looks like a well-known public figure” or “appears to show someone under 18”
Our own check, and it can't be confirmed past. Use a different picture, or describe an original character. A drawn, cartoon, anime or 3D-animation character of any age is fine: this check is about real-looking people. A glossy 3D render that could pass for a photo can be read as one, so keep the style clearly drawn.
What can set it off by mistake
Content filters are cautious, and they sometimes refuse work that is perfectly safe. These are the usual reasons:
- Swimwear, underwear or revealing clothing.
- Fitness, sports and close shots of the body.
- Medical, anatomical or injury pictures.
- Words for weapons, blood or violence, even in a harmless story.
- A close-up face that reads like a photo of a real person.
- A picture sent on its own, with no words to say what it is for.
What to do
- Reword the promptReplace the words that could be misread and describe the scene plainly: who or what is in the frame, the setting, the light.
- Change the pictureTry another reference, or a cleaner crop of the same one. A borderline picture can be refused even with a safe prompt.
- Say what a picture is forIf you sent only a picture, add a few words about what you want made from it.
- Try another modelEvery maker draws its line in a different place, so a prompt one model refuses can run on another. Compare them on the Models page.
- For a real person, use a registered likenessA person registered with their consent for your workspace can be used in video. Ask support to register them.
Credits
- A declined or stopped generation is never charged. Credits are taken only when a result is saved.
- While a generation runs, its price is held. When it is declined, the hold is released at once.
- Trying again is a new generation at its own price, shown before you press Generate.
What is never allowed
Some things can't be made on Plug-in AI however they are worded. They are listed in our Terms, under Acceptable use and Real people, faces and voices.
Still stuck?
Write to info@plug-in.ai with the generation's id (on its page in the Library) and we'll look into it.