Try this in the studio
If your Seedance prompt was rejected
- 1
Read the message on the job
It says whether the prompt, a reference, the sound or the finished clip was stopped.
- 2
Take out names and brands
Remove brand names, logos, celebrities and characters from films, games or TV. Describe the look instead.
- 3
Use a Character for faces
Save a photo you have permission to use as a Character, then call it by its @handle instead of uploading a raw photo.
- 4
Soften the wording, not the idea
Replace graphic or explicit phrasing with what the camera sees. Keep the scene, drop the shock words.
- 5
Turn off sound if the audio was flagged
Spoken lines and recognisable music can be stopped on their own. Generate without audio and add your own.
What Seedance stops, and what to do
Messages are the ones the KreatorFlow studio shows. Provider details as read on 15 November 2026.
| What was stopped | What to do |
|---|---|
| A prompt with explicit or policy-violating wording | BytePlus's input filter checks the prompt first. Rephrase it. The job is refunded. |
| A raw photo of a person used as a reference | Save it as a Character and use its @handle. Saved characters go through BytePlus's review for real-person references. |
| A saved character that looks like a real person, public figure or logo | The review refused it. Remove it or use a different one. |
| Brand names, logos, celebrities, famous scenes or songs | The result was flagged as possibly copyrighted. Describe the look without the names. |
| The generated audio | Turn off audio, or change what the prompt asks the audio to say. |
| The finished video | The model's own output check stopped it. Change the prompt and run it again. |
Sources, as checked:
- BytePlus: why was my request rejected, read 15 November 2026: byteplus.com/en/help/article/why-was-my-request-rejected-for-sensitive-or-non-compliant-content-and-how-do-i-reduce-false-rejections
Is Seedance censored?
Seedance is not uncensored, and no setting turns its checks off. Seedance runs on BytePlus, which screens the prompt and references before a clip is made and the finished video after it, and BytePlus says some of these platform safety policies cannot be disabled. On KreatorFlow, Pro, Max and Business renders go to a less filtered endpoint, so ordinary action, horror, fashion and realistic-people prompts are refused less often. Anything involving minors, sexual content of real people, non-consensual imagery or illegal acts is never allowed on any plan. If a normal prompt was refused, rephrase it, remove brand names and famous people, and save a face as a Character instead of uploading a raw photo. Blocked jobs are refunded.
BytePlus describes the setup as a two-layer safety system: an input content filter and the model's own output moderation. A request can be stopped before it starts, or the finished video can be stopped after it is made. KreatorFlow does not control either layer. We show the reason that comes back and refund the job.
What is never allowed, on any plan
These come from our Acceptable Use Policy. A less filtered route does not change them, and trying to get around moderation is itself against the policy.
- Sexual content involving minors, or anything that sexualises a minor.
- Non-consensual intimate imagery, sexual deepfakes, or sexualised content of a real person without consent.
- Illegal content and instructions for illegal acts.
- Harassment, threats, doxxing and defamatory claims.
- Fraud, phishing, impersonation and misleading commercial content.
- Content that infringes copyright, trademarks, privacy or publicity rights.
- Attempts to bypass moderation.
Seedance no restrictions, unrestricted and uncensored: what is true
BytePlus offers Seedance through more than one endpoint. On KreatorFlow, Pro, Max and Business video renders are sent to the less filtered one, and Free and Creator renders use the standard filtered one. That is the whole difference: fewer filters, not none.
Our Acceptable Use Policy says it plainly. Unrestricted workflows may have fewer creative limits, but provider and platform safety checks can still reject requests, especially where real people, privacy, protected likenesses, explicit sexual content or rights issues are involved.
So we do not sell Seedance as an NSFW or no-filter tool, and we cannot promise that adult content will pass. A site that says it can switch off the checks on a ByteDance model is not describing something BytePlus offers.
Why did Seedance reject my prompt?
Automatic safety checks block some prompts and references, especially anything involving minors, non-consensual imagery of real people, violence and copyrighted material. Blocked jobs are refunded.
Most refusals of ordinary prompts come from a few things: a real person's face or name, a brand or franchise, graphic wording, or sound that resembles a known song. The check is automatic and can be wrong, which is why the fixes below usually work for a legitimate idea.
How to fix a false rejection
Change one thing at a time so you can tell what the check disliked. BytePlus's own advice is to optimise the prompt to avoid explicit or policy-violating descriptions and to remove sensitive elements from input images, audio or video.
- Describe, do not name. Write "a red sports car with pop-up headlights", not a brand and model.
- No real people by name. Use a made-up character, or a Character saved from a photo you own.
- Say what the camera sees. Keep violence and injury implied, not graphic.
- Trim references. A clip with a visible face, a logo or a song can be refused even when your prompt is fine.
- Retry after one edit. A refused job is refunded, so trying again costs nothing extra.
Will another model be less strict?
Not as a rule. Wan and Kling in the studio have their own content checks too, and a refusal there reads "didn't pass the content check". Switching model can change what passes, but it is not a way around the rules above.
Do I lose credits when a prompt is blocked?
Yes. Failed and blocked generations are refunded automatically, usually within a few minutes.


