<a target="_blank" rel="ugc" href="https://www.instagram.com/jennifer.connelly/">Jennifer Connelly</a> 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 <a target="_blank" rel="ugc" href="https://ko-fi.com/jernaugurgeh22606">buy me a coffee. </a>:)Curious about my work process? I have summarized it <a target="_blank" rel="ugc" href="https://ko-fi.com/Post/My-TI-Training-Process-E1E1V8KLC">here</a>.Updated! Please check the "About this version" section for more info.<h2 id="heading-386">How to create a good prompt using my TIs</h2>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)"

Jennifer Connelly (80s/90s)

Retrained from scratch with a new dataset. This time I managed to get ahold of a few decent quality pics, and I think that shows in the final result. The embedding I'm uploading is step 1850. I trained at a 0.004 learning rate until step 1000. From there to step 2500, at a 0.001 rate. There were signs of saturation from very early on the training process, but by simply reducing the embedding weight in the prompt to 0.99 you will get rid of most of said saturation.

Files with SHA256: 72cfc241eed8f68191084019b7d7276599bf8c1f13b3ed3f2b15565e2e723c20