Use Guidelines

This document helps you understand the common reasons why video generation may fail, and how to adjust your image, file, or prompt before trying again.

If a generation fails, the credits used for the failed task are usually refunded automatically. You can check this in My Credits / Credit History. If you still do not see the refunded credits after a few minutes, please contact support and provide your account email and generation ID.

1. General Sensitive Content / Content Moderation Filter

Typical error messages / backend errors

You may see messages similar to:

• Content could not be processed
• Your request may violate content policy
• The input was rejected by safety system
• Sensitive content detected
• Moderation failed
• Safety checker failed

What this usually means

This usually means the system thinks your image, video description, or prompt may contain sensitive content. It does not always mean your input actually violates the rules; sometimes the model safety system may flag content by mistake.

Common triggers include:

• overly revealing, sexually suggestive, or adult content;
• gore, violence, weapons, self-harm, or dangerous behavior;
• minors combined with adult, violent, dangerous, or sensitive scenarios;
• real people, celebrities, public figures, or political figures;
• words that may be misread as body-related, intimate, medical, frightening, or threatening content;
• images that contain sensitive poses, nudity, weapons, blood, or extreme scenes.

How to adjust your uploaded content

You can try:

1. Remove language related to attacks, humiliation, threats, discrimination, hate, extremism, or illegal activity.
2. Avoid describing real harm, abuse, humiliation, coercion, intimidation, or high-risk events.
3. Replace sensitive descriptions with neutral, commercial, lifestyle, or creative wording.
4. If it is a story or ad scene, clearly state that it is safe, non-harmful, non-adult, and non-illegal.

Example revision

Not recommended:

A celebrity woman in a sexy outfit dancing seductively in a bedroom.

Recommended:

A fictional fashion model showcasing a stylish outfit in a bright studio, confident and elegant, commercial video style.

2. Public Figure / Celebrity / Political Figure Filter

Typical error messages / backend errors

• Public figure detected
• Celebrity or public person not allowed
• Face or identity safety restriction
• Request blocked due to public figure policy

What this usually means

If your prompt or image includes a celebrity, public figure, political figure, influencer, athlete, actor, or singer, the system may block the generation. Even if you only intend to imitate a style, it may still be detected as real-person generation or identity impersonation risk.

How to adjust your uploaded content

You can try:

1. Do not directly use the names of celebrities, public figures, political figures, or influencers.
2. Do not upload photos of public figures as reference images.
3. Do not ask the model to generate a real person speaking, endorsing, performing, or doing a specific action.
4. Do not ask the model to imitate a real person’s face, voice, identity, or public image.
5. Use fictional characters or general character descriptions instead.

Example revision

Not recommended:

Make Elon Musk introduce my product in a commercial video.

Recommended:

A fictional tech founder presents a new product in a clean startup-style commercial video.

3. NSFW / Adult / Sexually Suggestive Content

Typical error messages / backend errors

• NSFW content detected
• Sexual content blocked
• Explicit content not allowed
• Nudity detected
• Inappropriate content

What this usually means

This means the system detected possible nudity, adult content, sexual suggestion, provocative poses, sexual context, or descriptions that over-emphasize body parts in your image or prompt.

How to adjust your uploaded content

You can try:

1. Remove words related to nudity, pornography, adult acts, or strong sexual suggestion.
2. Avoid emphasizing body parts, provocative poses, intimate actions, or adult marketing language.
3. For fashion, beauty, or fitness ads, frame the content as product showcase, lifestyle, health, fashion, or commercial shooting.
4. Use safe, natural, non-revealing, non-adult reference images.

Example revision

Not recommended:

A seductive woman in revealing clothes poses on a bed.

Recommended:

A confident model wearing a summer outfit presents the product in a bright lifestyle studio.

4. Minor-Related Safety Filter

Typical error messages / backend errors

• Minor safety restriction
• Child safety policy
• Sensitive content involving minors
• Request blocked due to minor-related content

What this usually means

If your image or prompt involves children, teenagers, students, babies, or other minors, and the scene also contains adult, dangerous, violent, medical, body-related, emotional harm, intimate, or sensitive elements, the system may block the generation.

How to adjust your uploaded content

You can try:

1. Avoid placing minors in adult, dangerous, violent, horror, harmful, or sensitive scenarios.
2. Do not upload real photos of minors for high-risk or commercially misleading generation.
3. For children’s products, education, or family scenes, keep the scene safe, everyday, positive, and non-sensitive.
4. If needed, use adult characters, fictional characters, product-only shots, or general family scenes instead.

Example revision

Not recommended:

A scared child running away in a dark violent scene.

Recommended:

A child happily playing with educational toys in a bright and safe family room.

5. Violence / Weapons / Dangerous Behavior Filter

Typical error messages / backend errors

• Dangerous content detected
• Violence or weapon content blocked
• Self-harm or unsafe act restriction
• Criminal activity policy

What this usually means

The system may have detected violence, gore, weapons, explosions, self-harm, dangerous challenges, criminal activity, or content that could cause real-world harm.

How to adjust your uploaded content

You can try:

1. Remove descriptions of blood, wounds, death, severe injury, self-harm, crime, or weapon use.
2. Do not include specific steps for dangerous actions.
3. If it is an action, suspense, or game-style scene, describe it as “stylized, no gore, no realistic injury.”
4. If it is a product ad, make sure the product is used safely, legally, and in a normal consumer context.

Example revision

Not recommended:

A realistic scene showing how to attack someone with a knife.

Recommended:

A stylized fantasy action scene with dramatic lighting, no blood, no realistic injury.

6. Copyrighted Characters / Third-Party IP / Brand Imitation Filter

Typical error messages / backend errors

• Copyrighted character detected
• Third-party IP restriction
• Brand or character policy
• Request blocked due to copyrighted content

What this usually means

If your prompt or image includes movies, anime, games, comic characters, or asks the model to clearly imitate a protected character, fictional universe, or brand identity, the system may block the generation.

How to adjust your uploaded content

You can try:

1. Remove specific brand names, movie names, game names, anime names, character names, and logo names.
2. Do not upload movie screenshots, game footage, brand posters, copyrighted characters, or images/videos with logos.
3. Use original assets, self-shot materials, or materials you have permission to use.
4. Replace specific IP references with general style descriptions.

Example revision

Not recommended:

A Spider-Man style hero swings through New York.

Recommended:

An original masked superhero swings through a futuristic city skyline, cinematic action style.

7. Audio / Dialogue Generation Failure

Typical error messages / backend errors

• Audio generation failed
• Speech generation failed
• Audio moderation failed
• Dialogue generation failed
• Failed to synthesize audio

What this usually means

The main video content may be fine, but the audio, voiceover, dialogue, lyrics, music, sound effects, or uploaded audio may have triggered moderation or failed during generation.

Common causes include sensitive audio scripts, requests to imitate celebrity voices, copyrighted music, movie dialogue, poor audio quality, or temporary upstream audio service issues.

How to adjust your prompt

You can try:

1. Simplify the voiceover or dialogue, and remove sensitive, adult, violent, dangerous, or copyright-related content.
2. Do not ask for voices that imitate celebrities, public figures, or unauthorized real people.
3. Do not upload copyrighted songs, movie dialogue, third-party music, or branded sound effects.
4. Use clear, short, normal commercial voiceover lines.
5. If audio keeps failing, try generating a version without audio first, then handle audio separately.

Example revision

Not recommended:

Three people argue quickly for 20 seconds with overlapping dialogue about medical claims and financial guarantees.

Recommended:

A single spokesperson says one short line: “Create product videos faster with AI.” Clean studio background, friendly tone.

8. Temporary Upstream Service Error / Model Timeout / Queue Error

Typical error messages / backend errors

• Upstream timeout
• Model timeout
• Request timed out
• Generation took too long
• Provider timeout
• Internal service error
• Provider internal error
• Queue error
• Enqueue failed
• Failed to fetch result
• Temporary service issue

What this usually means

This is usually not caused by your prompt or image. It is more likely a temporary issue with the model provider, queue service, or generation service. The model may be busy, unstable, timed out, or the result may have failed to return.

How to handle it

You can try:

1. Retry the same task later.
2. If it fails repeatedly, try a different model.
3. Simplify your prompt or reduce complex materials before trying again.
4. Check that your uploaded materials are clear, accessible, and not corrupted.
5. If it fails multiple times, contact support and provide your generation ID, account email, and approximate generation time.
6. Credits for failed tasks are usually refunded automatically.

9. What should I do after a generation fails?

You can check in this order:

1. Review the failure reason shown on the page.
2. Confirm whether credits have been automatically refunded.
3. If it is a content moderation issue, remove sensitive words, public figures, copyrighted characters, adult content, or dangerous content.
4. If it is an upstream service issue, retry later or switch models.
5. If the same issue happens repeatedly, contact support.

10. What should I provide when contacting Support?

To help us diagnose the issue faster, please provide:

1. Your account email;
2. A screenshot of the failed generation page, including the prompt and error message.

This helps us determine whether the issue is caused by input materials, content moderation, or a temporary model service error.

11. Invalid Parameters / Uploaded Materials Do Not Meet Requirements

Typical error messages / backend errors

You may see a message such as:

Some input parameters are invalid or unsupported.
Please check file type, image size, prompt length, or other settings.

What this usually means

This usually means some input parameters submitted to the model do not meet the requirements. It does not necessarily mean the content violates policy.

Common causes include:

• The uploaded image format is not supported;
• The image is too small, too large, or its resolution does not meet the model requirements;
• The image is too blurry, heavily cropped, or the main subject is unclear;
• The uploaded file is corrupted, unreadable, or failed to load;
• Too many images were uploaded at once;
• The prompt is too long, too complex, or includes a format the model does not support;
• The selected duration, aspect ratio, model, or other settings are not compatible with the current input materials.

How to adjust your uploaded content

You can try:

1. Use common image formats such as JPG, JPEG, PNG, or WEBP.
2. Upload clear, complete images with an obvious subject. Avoid very small, blurry, heavily cropped, or screenshot-like materials.
3. If you uploaded multiple images, reduce the number of images and keep only the most important one or a few key references.
4. Simplify the prompt. Avoid very long scripts, repeated instructions, complex formatting, or too many unnecessary details.
5. If you changed multiple generation settings, restore the default settings first and try again.
6. For Product Ads / Photos scenarios, prioritize images that clearly show the product, person, or usage scene.

Example revision

Not recommended:

Upload several blurry screenshots with a very long prompt containing many unrelated instructions, multiple scenes, and complex camera movements.

Recommended:

Use one clear product image and a short prompt:
Create a short product ad video showing this product in a clean lifestyle scene, with natural camera movement and bright lighting.