Version 2 is released with vastly improved perspective compression. Z-image: between 0.5 and1.3 strength WAN 2.2 version is best at 2 for High, and 1 for Low.BEFORE AND AFTER DEMO VIDEO FOR WAN 2.2 Version:<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/TkqgsjeN180" start="0"></iframe></div>I'm a professional photographer and programmer with 20 years of experience across a wide range of lenses. While I love working with AI models, the lack of genuine lens control has always frustrated me — so I'm releasing my first lens LoRA. If there's interest, I'll create more.Note: The Wan and Z-Image versions use different trigger words — "telephoto" for Wan, and "600mm" for Z-Image.This LoRA replicates the look of a long prime lens. It was trained exclusively on material shot with a 600mm lens, and with the right prompting, you can achieve that level of compression. At other times, it may produce the feel of a 400mm, 200mm, 150mm, or similar — all of which are significantly harder to replicate without this LoRA.Tips: Prompting your background, middle ground, and foreground makes a real difference.To get the full long lens aesthetic, prompting for a full-length shot of a person helps considerably if your image is vertical — this will regularly deliver authentic long focal length perspective and full 600mm looks can be achieved.For subjects where a 600mm is rarely used, like in side a building, achieving the effect is trickier — you may end up with something closer to 150mm. It's still possible, but you'll need to refine your prompt to push as far as you can to the 600mm look.If you'd like to support me in creating more LoRAs from my lens collection, you can donate here: <a target="_blank" rel="ugc" href="https://buymeacoffee.com/lensloras">https://buymeacoffee.com/lensloras</a>

Long Prime Lens Lora (up to 600mm) (Wan 2.2 & Z-Image)

Low Model, Ise this model at 1.0 for best results Wan behaves very differently that Z-image with this data set, but still becomes vastly more cinematic, Using High Model is essential, and Low noise cleans up a lot of the details

Files with SHA256: ad888cdd0311a56f87cfac802790c772b8999d64bd21afa83293f719585f1b91