<h3 id="lora-description-ws0fm1epw">LoRa description</h3>These LoRa's can be called as "Photo Moodificators". It adds realism, depth and color / light grading. It gives quite good photographic quality even with empty negative prompt.This LoRa was made from HighRes travel photographs shoot with professional lens with shallow depth of field. No particular theme. Just a random moody photos from travels (nature, streets, animals, people, etc.).<h3 id="note-on-lora-rank-4iymks9zb">Note on LoRa Rank</h3>Rank 128 in some tests showed lower quality / performance. But Rank 256 is not 2-times better than Rank 128. So, I decided to release both ranks for you to test. You can start with R128.<h3 id="prompting-xu78nkuuz">Prompting</h3>I've used a mixed approach to caption the images in dataset. The captions in Training Data (<a target="_blank" rel="ugc" href="http://captions.zip">captions.zip</a>) contains the mix of WD tags and natural description made by LLM 7b (Llama Mistral 7b).So, the best advice on getting idea about prompt words is to look into the captions and the example images I am posting here.Part 1 - first 500 captions.Part 2 - next 500 captions.Part 3 - last 200-300 captions.PS: sorry... no images in <a target="_blank" rel="ugc" href="http://captions.zip">captions.zip</a>. I feel very shy about some of my photographs.

Travel Photography

Merged Parts 1..3 into one LoRa with the same rank 128.

travel_photography_dim128.safetensors

captions.zip

Files with SHA256: 1c9fbfc41ee6eca4066a2cb433f024c495ed2b4e009825533bf8eed4a44f3633