Lily Chloe Ninette Thomson, known professionally as Lily James, is an English actress. Born: 5 April 1989
Fuel my GPU's caffeine addiction! (Ko-fi)
https://ko-fi.com/cyberaimania
Trained on Wan 2.1 T2V-14B model locally on RTX 4090 with 125 publicly accessible images using diffusion-pipe. Training data and config.toml available for download.
Trigger: jamslix
Example prompts:
--
Medium close-up shot, jamslix woman, initially seen in slight profile talking on an old Nokia phone by a window in a slightly cluttered living room. She then turns her head to look directly at the camera, face clearly visible. Natural daylight streams in. Slight camera shake. Early 2000s home video aesthetic, slight film grain, realistic.
--
Medium long shot, female jamslix, with her face visible, confidently walking down a red carpet lined with photographers at a movie premiere. She wears a stunning evening gown and smiles dazzlingly at flashing cameras. Cinematic lighting, slow motion, elegant atmosphere, high detail, photorealistic.
--
Over-the-shoulder shot, focusing on jamslix, a 30-year-old woman, face visible in profile as she carefully waters houseplants on a windowsill. Background shows a typical room from the early 2000s. Handheld camcorder feel, natural lighting, slightly desaturated colors, VHS aesthetic.
--
I would really appreciate it if you could rate and comment on my lores. Your feedback, whether positive or constructive criticism, truly gives me a "kick" and the energy to create new projects. Thank you in advance for your engagement!
Check out my other LoRAs:
Claire Forlani:
https://civarchive.com/models/1465482/claire-forlani-or-wan21-t2v-14b
Zendaya:
https://civarchive.com/models/1469052/zendaya-coleman-or-wan21-t2v-14b
Hayley Elizabeth:
https://civarchive.com/models/1473023/hayley-elizabeth-or-wan21-t2v-14b
Enjoy!
Description
FAQ
Comments (3)
You read my mind! I had just added her to my list of characters to train yesterday.
Thanks for doing this!
Thanks! You're doing an amazing job too!! I see you successfully completed training on VIDEO — well done! I wanted to try it myself, but I wasn't sure if 256 x 256 would be a good choice. I was worried about the final quality of the LoRA.
I tried fitting 480 x 480 clips into the VRAM of our RTX 4090 XD, but it was only possible with a very high
blocks_to_swap = 30 - 38
and
gradient_accumulation_steps = 8,
which unfortunately would have taken forever on our GPU.
So I gave up on training at 256 x 256, but I can see now that I shouldn’t have — your LoRA turned out fantastic!
--
I just finished training for Scarlett Johansson — I’ll probably upload it tomorrow. Just letting you know so we don’t end up making duplicates haha. Cheers, mate!
@CyberAImania Thanks for the props! Yep, 256x256 was sufficient and surprisingly fast to train!
Looking forward to more models from you.