For research and educational purposes only. Use responsibly.

Try it 100% free on Discord

The only SynthID cleaner that actually works.

Join Discord - Try Free
01

Subscription Tiers

04

Free

$0
  • Credits 10 / month
  • Concurrent Jobs 1
  • Low-Priority Cost 3 – 5 credits
Join Discord
03

Plus

$9/mo
  • Credits 150 / month
  • Concurrent Jobs 2
  • Low-Priority Cost 1 – 3 credits
Subscribe
02

Pro Popular

$29/mo
  • Credits 500 / month
  • Concurrent Jobs 4
  • Low-Priority Cost 0 – 2 credits
Subscribe
01

Ultra

$99/mo
  • Credits 2,000 / month
  • Concurrent Jobs 8
  • Low-Priority Cost FREE
Subscribe
02

Credit Packs

One-time purchase. Need more credits? Buy a pack anytime.

20
Credits
$2
$0.10 / credit
120
Credits
$10
$0.083 / credit
Best Value
400
Credits
$30
$0.075 / credit
1,500
Credits
$100
$0.067 / credit
Buy Credits
03

How to Use

Get started in under a minute. No downloads, no setup.

Getting Started

  1. Go to a generation channel (#low-priority)
  2. Use the /i2i command
  3. Attach your image
  4. Adjust denoise and submit

Denoise Settings

0.15 Subtle — lower chance of bypass
0.3 More change — higher chance of bypass

Use the default first!

Generation Costs

High-Priority

3 – 5 credits (based on resolution)

Low-Priority (discounted for paid tiers)

TierLow-Priority Cost
Free3 – 5 credits
Plus1 – 3 credits
Pro0 – 2 credits
UltraFREE

Commands

/i2i Generate an image (bypass SynthID)
/credits Check your balance & tier

Tips

  • Lower resolution = fewer credits
  • Images with large faces may need higher denoise to pass
  • Buy credit packs if you run out!
04

Open Source Research

AI safety research showing a working bypass to Google's SynthID on Nano Banana Pro.

00quebec / Synthid-Bypass
300
47
01

Low-Denoise Regeneration

Very low denoise values in KSampler "repaint" images, replacing watermark-carrying pixel noise while keeping visual content intact.

02

Multiple Passes

Sequential KSamplers iteratively scrub the watermark away through carefully configured multi-stage diffusion pipelines.

03

Structural Preservation

ControlNets with Canny edge detection and face restoration maintain original composition and facial structure throughout processing.

04

Two Workflows

General-purpose pipeline for any image, plus a portrait-specialized workflow with YOLO face detection and targeted inpainting.

View on GitHub