Sierra Qwen

Wan2.2 T2V LoRA

why this thing exist: because she is beautiful. I wouldn’t die for any flag, country, religion, or moral principle, but I would gladly die for a beautiful Caucasian woman if all black women vanished from the earth. Skip the rest it's mostly brain fart.

Second this LoRA was made after I trained a Wan2.2 T2V LoRA (based on this character, which I’ll upload here later).

The main goal was to find a better way to generate longer videos. Right now, most current solutions for long videos are Wan2.1-based (like LongCat) and still have noticeable issues , especially hand and finger deformities.

Quality also tends to drop after around 15 seconds when just extending from the last frame. (I’ve had decent results with SmoothMix I2V v1, sometimes reaching 15 seconds without major hallucinations.)

Achievable for regular joes:

• If the T2V LoRA keeps the main character consistent ( 85% similarity),

• --> Use the T2V LoRA for full-body shots and distant scenes.

• --> Use I2V for close-ups.

This combination allows us to create much longer and more consistent videos while avoiding the common problems found in most current Wan2.1-based solutions.

(note: The SDXL version of this LoRA was released earlier, but the account got blocked by Civitai leftists and jihadists, It’s good idea to save a copy in case this account gets banned too.