German Suplex

This time, I created a wrestling move — the German Suplex.
It was trained from various angles — front, back, and side — and the success rate appears to be quite high.

The side view works best, and the front view (facing the viewer) also performs well, but for the back view (fully from behind), the prompt must be written carefully.

If the behind doesn’t generate properly, adding g@s@p@b to the prompt increases the success rate.
※ Use this special prompt only for back view shots (fully from behind). Like This.

---

Prompt Writing Guide

The attacker is referred to as attacker / and the victim as victim.

Example – Front:

A bat-man attacker appears quickly from above and goes behind the blonde victim, tightly locking their waist.
The victim opens her mouth wide in shock, her face showing a look of panic as she’s grabbed.
With explosive power, the attacker then lifts and launches the victim backward with a high German suplex, knocking her to the ground.
The victim flips completely overhead before crashing violently onto the canvas, landing squarely on her upper back and shoulders.
The attacker maintains the bridge perfectly, showcasing complete control of the impact.

Example – Side:

The blonde attacker on the right is holding the brown-haired victim on the left firmly with both arms.
With explosive power, she lifts the victim high into the air, arching her back all the way before slamming her down, driving the victim’s shoulders and the back of her head hard into the mat.

Example – Behind (back view shots)****:

g@s@p@b. The black-jacketed attacker on the left moves behind the beige-clothed victim on the right and holds her by the waist.
The attacker hoists the victim high into the air, arching her back completely, and drives the victim’s shoulders and head hard into the ground right in front of the camera with a German suplex.
As the victim crashes into the ground, dust bursts up and her crotch becomes visible.

---

Noise Settings

• High_noise was tested at 1.0.

• Low_noise was tested between 0.7–1.0.
  When the low_noise value is too high, outfit variations may occur, so around 0.8 is recommended.

The 81-frame motion is slightly slower, and as the frame count decreases, the motion becomes faster.

For detailed prompts, paste the uploaded video sample into ComfyUI—the app will generate a simple workflow and prompt you can refer to.