I’ve trained this LoRA based entirely on the official screenshots released by the DOOM: The Dark Ages team. To go further, I wrote a quick Python script that extracted high-res stills from the trailer — frame by frame — which I carefully selected and annotated for style consistency. It was time-consuming, but the quality of the frames was worth it: massive resolution, crisp details, and lots of variation in tone and lighting.The training ran locally and took quite a while — over 10 hours — so I stopped after the 6th epoch out of 10. Despite that, I’m really satisfied with the results and how well the style came through.The trigger word is "do2025om style". I've had the best results with a fixed CFG of 2.5, with euler as sampler with normal or simple scheduler, with a LoRA strength between 0.85 and 1, but feel free to experience things and test new stuff!If you like my work you can follow me on my twitter that I just created, planning to release more content there!<a target="_blank" rel="ugc" href="https://x.com/AiSaucyvisuals">👨‍🍳 Saucy Visuals (@AiSaucyvisuals) / X</a>You can also buy me a kofi here: <a target="_blank" rel="ugc" href="https://ko-fi.com/saucyvisuals">Support 👨‍🍳 Saucy Visuals</a>Hope you will cook some nices images 👨‍🍳

Files with SHA256: 57e800e834e9bf8cb1f832ab0d2b5ebe40be5de76d64e106af924b696cca14d8