Merged the Kroma v0.1 lora by Lodestones into Krea 2 Turbo BF16 model and quantized it to a mix of INT8 Convrot & W4A4 layers.

Using loras normally through ComfyUI slows down my generation speed by 60-70% so did this instead.

Once again, just the first stage (T2V). I'm too lazy for second stage, upscaling and what not.

Tested on:

• NVIDIA GTX 1660 Super 6 GB

• 32 GB System RAM

Test settings:

• 8 steps

• CFG 1

• Euler / Simple

• 696 × 1048 resolution

• Approximately 8 s/it on a GTX 1660 Super