How to Bypass Seedance 2.0 Face Detection
Your character reference keeps getting rejected? Here's the grid overlay trick that bypasses the face filter in 10 seconds. Fair warning: the grid needs to be solid, and the generated video will likely show grid artifacts — but for many use cases, it's a worthwhile trade-off.
Why Seedance 2.0 Blocks Your Images
You upload a perfectly fine character reference — your own face, a stock photo, even a cartoon — and Seedance 2.0 rejects it. “Face detected.” Here's what's happening under the hood.
Content moderation
Seedance 2.0 uses automated face detection to flag uploads that contain recognizable human faces. This is a safety measure designed to prevent deepfakes and unauthorized use of someone's likeness.
False positives are common
The detector isn't perfect. It can flag cartoon characters, stylized illustrations, side profiles, partially obscured faces, and even images where the face is small or far away. Legitimate creative use cases get caught in the filter constantly.
No manual review option
There's no way to submit an appeal or request manual review. If the system detects a face, your image is rejected — period. You need a technical workaround.
4 Ways to Get Around Face Detection
We tested multiple approaches. The grid overlay is the clear winner for reliability and quality, but here's the full comparison so you can decide for yourself.
Grid Overlay (Recommended)
Best methodDifficulty: Easy · Effectiveness: HighAdd solid grid lines over your image. The grid breaks up the pixel patterns that face detection relies on, so Seedance 2.0 still understands the character, pose, and scene. The grid needs to be solid (100% opacity) to reliably fool the detector — semi-transparent lines usually aren't enough. The trade-off: the generated video will likely show visible grid line artifacts.
- +Works consistently across different face types and angles
- +Free tool available — no software installation needed
- +Takes 10 seconds per image
- +Character identity and pose still come through clearly
- −Generated video will likely show visible grid line artifacts — this is the main trade-off
- −Solid lines are needed to reliably bypass detection, so you can't just make the grid invisible
- −Finding the right grid density (sparse enough for quality, dense enough to bypass) takes experimentation
Cropping / Partial Face
Difficulty: Easy · Effectiveness: MediumCrop the image so only part of the face is visible — forehead to nose, or nose to chin. With less than ~60% of the face visible, the detector often fails to trigger.
- +No additional tools needed
- +Works well for close-up character references
- −Loses full-face context — Seedance may invent the missing parts differently each generation
- −Inconsistent results for side profiles
- −Not viable when you need full-body character references
Downscaling + Re-uploading
Difficulty: Easy · Effectiveness: Low–MediumReduce the image resolution significantly (e.g., to 256×256), then re-upload. Lower resolution means fewer pixel details for the detector to latch onto.
- +Dead simple — any image editor works
- −Lower-res inputs = lower-quality character reference for Seedance 2.0
- −Doesn't work reliably for clear, front-facing portraits
- −Quality trade-off may not be worth it for final renders
Style Transfer Pre-processing
Difficulty: Medium · Effectiveness: MediumRun the image through a style transfer filter (oil painting, watercolor, comic book) before uploading. This changes the pixel-level structure enough to dodge detection while preserving the overall character appearance.
- +Can produce interesting artistic results
- +Works well when your target video style is already stylized
- −Adds an extra step — need a style transfer tool
- −Realistic character references get distorted
- −Seedance may pick up the style filter as the intended art style
The Grid Overlay Method — 5 Steps
This takes under a minute. We built a free grid overlay tool specifically for this — it runs in your browser, nothing gets uploaded to any server.
Open the Grid Overlay Tool
Go to the free Grid Overlay Tool. No signup, no download — it runs entirely in your browser. Your images never leave your device.
Upload your reference image
Drag and drop or click to upload the character reference image that's getting blocked by face detection. Supports JPG, PNG, and WebP.
Adjust the grid settings
Start with the default 3×3 grid at full opacity (100%), white, 10px line width. If face detection still triggers, increase to 6×6 or 10×10. The lines need to be solid — semi-transparent grids usually don't fool the detector. Sparser grids (3×3) leave more of the image clean but may not bypass detection on clear portraits; denser grids (10×10+) are more reliable but leave more visible artifacts in the output.
Download the processed image
Hit the download button. You'll get a PNG with the grid overlay baked in. The file size will be similar to the original.
Upload to Seedance 2.0
Use the processed image as your character reference in Seedance 2.0. Tag it with @Image1 and describe its role (e.g., "@Image1 is the main character"). The face detection should no longer trigger. Note: the generated video will likely show grid line artifacts where the overlay was — this is the expected trade-off.
Recommended Grid Settings
Different images need different grid density. The key rule: keep opacity at 100% (solid lines) — semi-transparent grids don't reliably bypass Seedance 2.0's face detection. The trade-off is between grid density and output artifact visibility.
Light 4×4
Grid: 4 × 4
Line width: 15
Opacity: 100%
Color: White
Fewer lines = fewer artifacts in the output. Good starting point, but may not bypass detection on very clear portraits.
Standard 6×6
Grid: 6 × 6
Line width: 12
Opacity: 100%
Color: White
The recommended preset for most images. Reliable bypass with moderate grid artifacts — works well for drafts and social media content.
Dense 10×10
Grid: 10 × 10
Line width: 8
Opacity: 100%
Color: White
Most reliable bypass, but the output will have more visible grid line artifacts. Use when sparser grids fail on clear, well-lit portraits.
Common Issues & Fixes
Grid overlay applied but face detection still triggers
Fix: Make sure opacity is at 100% (solid lines) — semi-transparent grids usually don't work. If it still triggers, increase grid density (try 6×6 or 10×10) or increase line width. For very clear, high-resolution portraits, you may need a denser grid. You can also combine with a slight downscale to 1024px width.
Grid lines visible in the generated video
Fix: Try adding negative prompt cues like "no grid lines, no overlay, no mesh, clean skin, smooth image" to your prompt. This can help Seedance 2.0 suppress grid artifacts during generation. Also try using a sparser grid (3×3 or 4×4) with thinner line width to reduce the source overlay. Combining prompt guidance with lighter grids often produces noticeably cleaner results.
Character looks different from the reference
Fix: This is usually a prompt issue, not a grid issue. Make sure you're using @tags to explicitly assign the image as a character reference (e.g., "@Image1 is the main character, a woman in her 30s with dark hair"). Adding descriptive details in the prompt helps Seedance lock onto the right features.
Works for some images but not others
Fix: Face detection sensitivity varies by image. Front-facing, well-lit portraits with clear eye contact are hardest to bypass. Try a 3/4 angle reference photo instead, or use a denser grid for those specific images.
Maximize Your Character Reference
Getting past face detection is step one. Getting a great character-consistent video is step two. Here are the prompting patterns that work best with grid-overlaid references.
Always use @tags
Don't just upload — assign a role. “@Image1 is the main character” tells Seedance 2.0 exactly what to do with the reference. Without a tag, it might use the image as a background or style reference instead.
Describe the character in the prompt too
Don't rely solely on the image. Add key details in text: “@Image1 is the main character, a man in his 40s with short gray hair and a blue jacket.” This gives the model two anchors — visual and textual.
Use front-facing headshots
Character consistency is best with clear, well-lit, front-facing portraits. Side profiles and group photos give the model less to lock onto. Yes, front-facing triggers face detection more — that's exactly why the grid overlay tool exists.
Draft with Fast, render with Standard
Use Seedance 2.0 Fast for quick iterations — half the credits, same face detection behavior. Once the character looks right, switch to standard Seedance 2.0 for the final render.
Use prompt cues to suppress grid artifacts
If the output shows grid lines, add negative cues to your prompt: “no grid lines, no overlay, no mesh, clean skin, smooth image.” This guides Seedance 2.0 to suppress the grid pattern during generation. It doesn't eliminate artifacts 100% of the time, but it often makes a noticeable difference — especially when combined with a sparser grid (3×3 or 4×4).
For a complete guide to prompting, @tags, and workflow optimization, check out the full Seedance 2.0 tutorial.
Frequently Asked Questions
Why does Seedance 2.0 have face detection in the first place?
It's a content safety measure to prevent deepfakes and unauthorized use of real people's likenesses. The system automatically scans uploaded images for human faces and blocks them if detected. While well-intentioned, it frequently blocks legitimate creative use cases like using your own face, stock photos, or fictional characters.
Does the grid overlay affect video quality?
It can. The grid needs to be solid (100% opacity) to reliably bypass face detection, so the generated video may show grid line artifacts. However, you can reduce this by adding prompt cues like "no grid lines, no overlay, clean skin, smooth image" — this helps Seedance 2.0 suppress the artifacts during generation. Combining prompt guidance with a sparser grid (3×3 or 4×4) often gives noticeably cleaner results. It takes some experimentation, but it's workable.
Is this method against Seedance 2.0's terms of service?
The grid overlay technique modifies your own image before upload — it doesn't hack or exploit any API vulnerability. You're simply preprocessing your image, similar to resizing or color-correcting it. That said, always use this responsibly: don't use it to create deepfakes or generate content with real people's faces without their consent.
Can I use this for commercial projects?
Yes, if you have the rights to the original image. The grid overlay doesn't change the copyright status of your source material. If you're using your own photos, stock images with commercial licenses, or AI-generated character references, you're fine.
What grid settings work best?
Start with a 3×3 white grid at 100% opacity and 10px line width — this is the tool's default. If face detection still triggers, increase to 6×6 or 10×10. Keep opacity at 100% (solid) — lowering it usually causes detection to succeed again. You can reduce line width to minimize artifact visibility, but the lines need to be opaque. It's a balance between 'enough grid to bypass' and 'as little grid as possible for cleaner output.'
Does this work with Seedance 2.0 Fast too?
Yes. Both Seedance 2.0 and Seedance 2.0 Fast use the same face detection pipeline. The grid overlay bypass works identically for both variants.
My image doesn't have a real human face — why is it still getting blocked?
The face detector has a high false-positive rate for stylized illustrations, anime characters, 3D renders, and even profile views. If it looks vaguely face-shaped to the algorithm, it gets flagged. The grid overlay works on these false positives too.
Can I batch-process multiple images?
The current grid overlay tool processes one image at a time. For batch processing, you'd need to use a local script (ImageMagick or Python with PIL can add grid overlays in a loop). But for most workflows, processing 2–3 reference images takes under a minute with the online tool.
Showcase
AI-Generated Videos
Discover stunning AI-generated videos. From photography to architecture, people to creative projects — see how AI transforms ideas into cinematic videos.
Stop Fighting the Filter. Start Creating.
Apply the grid overlay, upload your character reference, and let Seedance 2.0 do what it does best.