Built for AI filmmakers — try FramesForge free
↓ Seedance 2.0 Utility

Seedance
Toolkit

Two tools in one: convert your frames into tiled reference panels that bypass Seedance's face detection — choose 3×3 for wide/medium shots or 5×5 for closeups and front-on mid-shots — and clean your prompts to pass the content filter. Free, no account required.

Gemini API Key — used for both tools
Stored locally in your browser only — never sent anywhere except Google's API.
⚠
Face detection fires before your prompt is read. Seedance 2.0 scans uploaded images for recognisable faces and rejects them silently — no prompt rewrite will fix this. The panel technique distributes the face across a grid so Seedance reads it as a texture or pattern rather than a person. Use 3×3 for wide or medium shots. Use 5×5 for front-facing mid-shots and closeups — Seedance's detector can still resolve a face in a 3×3 at close range. Generate your panel below, then upload it to Seedance instead of the original.
Frames — drop your keyframes here
⬚
Start Frame
Click or drag & drop
Optional
⬚
End Frame
Click or drag & drop
Panel Size
3×3 — 9 tiles. Works for most shots.
How to use the panels in Seedance 2.0
01
Generate your panel(s) below and download them.
02
Upload the panel image to Seedance as your reference — not the original frame. It will appear as @image 1.
03
Start your prompt with exactly: "Use character @image 1" — then your shot description follows. Do not include a separate character description block. Mentioning a character reference separately will cause the generation to fail.
04
For start + end frame control, upload both panels and begin with: "Use character @image 1 as the first frame and @image 2 as the last frame".
Generated Panels
⚠
Seedance's content filter is tightening fast under MPA pressure. Words like fight, attack, criminal, abduction, kill, brutal, blood trigger silent rejections — even in clearly cinematic contexts. This tool rewrites your prompt in director-language that carries the same intent but reads as shot description rather than incident narration.
Your prompt — paste it in
✓ Cleanser Output
Image filter — what to know
↑
If your generation keeps failing after a clean prompt, the image is failing, not the prompt. Seedance's face scanner runs before the prompt filter.
↑
Switch to the Panel Generator tab and convert your reference frame into a 3×3 panel. Upload that panel instead.
✕ Gets blocked

Uploading a direct photo or keyframe where a face is clearly visible and centred — especially front-facing mid-shots and closeups.

Seedance's face detector fires immediately — the generation fails silently before reading your prompt.

✓ Passes filter

The same image tiled in a panel grid. The face detection heuristic reads it as a repeating texture, not a person.

3×3 works for most shots. For front-facing mid-shots and closeups, use 5×5 — the detector can still resolve a face in a 3×3 at close range.

Want the full AI filmmaking pipeline?
FramesForge handles generation, masked editing, character consistency, shot sequences, and prompt construction — all in one tool. 30 free generations, no subscription.
Try FramesForge →