Jennifer Connelly is an American actress, known for many roles in films such as Labyrinth, Career Opportunities, Hulk and Requiem for a Dream, among many others. She won an Oscar for Best Supporting Actress for her role in A Beautiful Mind (2001).
Appreciate my work? My TIs are free, but you can always buy me a coffee. :)
Curious about my work process? I have summarized it here.
Updated! Please check the "About this version" section for more info.
How to create a good prompt using my TIs
You're obviously free to experiment, but bear in mind that my TIs are trained with a more or less fixed phrasing, that normally starts with:
"photo of EMBEDDING_NAME, a woman"
So I recommend always starting your prompt like that and then building the rest of the prompt from there. For instance, "photo of beautiful (jg_jc0n:0.99), a woman, (sfw tight fashion shirt:1.2), (long jeans), in a departmentn store, natural light, (masterpiece:1.2) (photorealistic:1.2) (best quality) (detailed skin:1.2) (intricate details) (8k) (HDR) (cinematic lighting) (sharp focus), (looking at the camera:1.1), (closeup portrait:1.1), (painted lips)"
Description
Retrained from scratch for 130 steps with with these settings.
FAQ
Comments (23)
quality work!!, as usual :)
Thanks, mate!
Nice :) Are they flagging everything for moderator review now for real people? I posted a couple images and even the ones with all grey tags are flagged for review. Weird.
Damn, I missed this comment the other day. I'm not sure what's been happening with the images, but I couldn't see many even if they were completely sfw and had no "color" tags attached, just the gray ones. Fortunately they eventually become visible, but sometimes it takes a bit too long...
Tutorial for this model Jennifer Connelly 80s and 90s easy and fast
Excellent tutorial! :)
Great idea but I don't really like the idea that the ti comes prepacked with the notebook. It takes away from the ti creator. I would prefer if the tutorial shows how to download the ti here from civit and how it can be used with that provided notebook. Link to the profile, show them how to download, where to move the files and how to use them.
@ElizaPottinger yeah, but easy and fast ;D, my graphic card is 4gb, and it takes forever to load models, that's why I started doing these tutorials, but when I can make a more powerful graphic I will do what you suggest.
It's gone!
Thanks for the great work.
I wonder if V3.0 is for COLAB. I could NOT get JC's face for my WebUI, but I can get her face just fine with V2.1. Would you please upload V3.0 working for WebUI?
Hi seejay! I'm honestly not sure wat you mean. All my files are generated as .pt files with Automatic1111 and should work on any web UI based on SD. Have you tried this tutorial?:
@JernauGurgeh Yes, I watched the https://youtu.be/1HZy7lqTDfg. But I did not use COLAB but Automatic1111. I expected that COLAB and Automatic1111 would show different results with the same settings, as it was mentioned in https://www.reddit.com/r/StableDiffusion/comments/10ugbhn/deforum_different_results_on_colab_and/, but not too much. Unfortunately V3.0 gave me someone very different from JC. The file V2.1 was okay. That's why I thought maybe V3.0 only works in COLAB. The big difference in file sizes(~ 7k for V3.0 and ~ 13k for V2.1) also made me think that the two versions are quite different. Do you think that V3.0 should work in local Automatic1111 just as in COLAB?
@seejay333 No, no, I never use COLAB for generating pics, (I always use Automatic1111 too), and I guarantee you can obtain the very same results as the ones I posted if you use exactly the same settings. The most usual causes for obtaining different results are the following:
- Different model (most of the pics in my examples were generated with AvalonTruVision)
- Different CLIP Skip (this happens inadvertently very often, as many people always use CLIP Skip 2, but for this TI and for basically all of the TIs I create CLIP Skip 1 should be used)
- No VAE or different VAE selected (vae-ft-mse-840000-ema-pruned.ckpt should be used)
- Other settings like LoRAs, ControlNet or whatever being used (I used none for generating the pics in the showcase)
It should work, mate. In case it doesn't for you, you can find me on Discord (zakalwe4319 is my username) and I'll try to help you. Good luck!
@JernauGurgeh I followed your instruction, but the result is the same as I described before. I will try to contact you through Discord some time later. Thank you much for your kind concern.
@JernauGurgeh For someone who would be interested in our communication, JernauGurgeh found my mistake. The TI name of V3.0 is jg-jc0n NOT jg_jc0n. Then it beautifully works. I appreciate the help a lot.
Any chance of getting an SDXL 1.0 version? Everything the SDXL models are generating for me looks like it's from Jennifer circa Snowpiercer.
Not for now, I'm afraid. I guess I'll eventually start working on XL, provided I upgrade my PC at some point, but for now my focus is solely on 1.5.
@JernauGurgeh Excuse me, but they said old TI is compatible with SD XL. Only v1.5 LORA are not.
@settima_ai No idea, mate. I don't use XL at all, so I can't say. As far as I know, they're not compatible, but maybe I'm wrong.
@JernauGurgeh May be it's time to try! You would suprised how is easy to use SD XL, no need to install, change or activated anything! Just download a XL checkpoint as you always do and put in the same folder you use for 1.5. Then select it as usual and... tah dah! You are a SD XL user now!
I suggest to start with a XL TURBO model, which is not so much slower than a SD 1.5:
https://civitai.com/models/112902/dreamshaper-xl
The only problem, it will ignore any LORA reference, but is much smarter with your pure prompt.
Then, it would be interesting to discover if your TI is working or not.
(by the way, I'm not a XL user too because I still have to finish some works with sd 1.5. I'm using Auto1111 with online Google Colab )
@settima_ai I think I actually gave it a try some months ago and didn't enjoy the experience, but perhaps with a turbo model it'll be different, as you say. Thanks for the tip! :)
So a TI can learn the body proportions, not only the face...
It definitely can. Depends on the dataset, of course, but the AI can at the very least learn the general proportions of the subject's body.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.