# malcolmrey's articles A collection of guides, tutorials, and resources about AI model training and generation, primarily focused on Stable Diffusion, SDXL, and Flux. **Note:** All articles listed below are hosted on [CivitAI](https://civitai.com/user/malcolmrey/articles). --- ## Training Guides ### [DreamBooth / LyCORIS / LORA Guide](https://civitai.com/articles/7/dreambooth-lycoris-lora-guide) **Comprehensive guide on training DreamBooth models and extracting LyCORIS/LoRA from them.** This tutorial covers the complete workflow from dataset preparation to model extraction, explaining how to train a Dreambooth checkpoint that can be converted to both LORA and LyCORIS formats. Includes detailed scripts and examples for the entire process. 🔗 [Read Article on CivitAI](https://civitai.com/articles/7/dreambooth-lycoris-lora-guide) ### [SDXL LoRA Training](https://civitai.com/articles/1591/sdxl-lora-training) **Step-by-step guide for training SDXL LoRA models using RunPod.** Covers the complete training flow for SDXL, including optimal parameters, dataset preparation, and cloud GPU usage. Discusses the importance of trigger word selection and provides detailed configuration examples for achieving high-quality results with SDXL models. 🔗 [Read Article on CivitAI](https://civitai.com/articles/1591/sdxl-lora-training) ### [Textual Inversion / Embedding Training Guide](https://civitai.com/articles/3114/textual-inversion-embedding-training-guide) **In-depth tutorial on training Textual Inversion embeddings using Kohya SS.** Explains how to create efficient 8KB embedding files that work across different base models. Covers training parameters, dataset requirements, and how embeddings differ from LoRAs and LyCORIS in terms of functionality and use cases. 🔗 [Read Article on CivitAI](https://civitai.com/articles/3114/textual-inversion-embedding-training-guide) ### [Flux Guide, Part I - LoRA Training](https://civitai.com/articles/9360/flux-guide-part-i-lora-training) **Complete guide for training Flux LoRA models with optimal settings.** Details the training process for Flux models, including setup with kohya_ss, dataset preparation, and training parameters. Covers resolution settings, step counts, and troubleshooting common issues. Based on 150+ Flux training experiments. 🔗 [Read Article on CivitAI](https://civitai.com/articles/9360/flux-guide-part-i-lora-training) ### [WAN 2.1 LoRA Training Tutorial](tutorials/wan-21-lora-training.md) **Complete guide for training WAN 2.1 LoRA models using AI Toolkit.** This guide, originally published on CivitAI, covers the process of training WAN 2.1 LoRA models with optimal dataset preparation (20 images, 2500 steps), step-by-step AI Toolkit setup instructions, and comprehensive best practices for image selection and training. Includes 130+ pre-trained LoRA models and ComfyUI workflow. � [Read Tutorial](tutorials/wan-21-lora-training.md) --- ## Generation & Quality Guides ### [Improving Results by Using Multiple Models of the Same Concept (Turning it to 11!)](https://civitai.com/articles/1721/improving-results-by-using-multiple-models-of-the-same-concept-turning-it-to-11) **Advanced technique for combining multiple LoRA and LyCORIS models to achieve superior likeness and consistency.** Explains how mixing different model types (LoRA + LyCORIS) at specific weights produces better results than using a single model. Includes weight recommendations and practical examples for achieving photorealistic quality. 🔗 [Read Article on CivitAI](https://civitai.com/articles/1721/improving-results-by-using-multiple-models-of-the-same-concept-turning-it-to-11) ### [Bringing it up to Twelve! Going Deep into Quality](https://civitai.com/articles/3527/bringing-it-up-to-twelve-going-deep-into-quality) **Advanced quality optimization guide combining LoRAs, LyCORIS, and Embeddings.** Builds on the "Turning it to 11" article by adding Textual Inversion embeddings to the mix. Covers optimal weight combinations, tips for eyes and skin quality, and explains how to use multiple models without overtraining artifacts. 🔗 [Read Article on CivitAI](https://civitai.com/articles/3527/bringing-it-up-to-twelve-going-deep-into-quality) ### [ADetailer and a Poll, and a Few Words About Roop](https://civitai.com/articles/1633/adetailer-and-a-poll-and-a-few-words-about-roop) **Guide on using ADetailer extension for automatic face enhancement.** Explains how to use ADetailer as an automated inpainting solution to improve face quality in full-body shots and distant subjects. Includes discussion of Roop face-swapping technology and best practices for combining different enhancement techniques. 🔗 [Read Article on CivitAI](https://civitai.com/articles/1633/adetailer-and-a-poll-and-a-few-words-about-roop) --- ## Tools & Scripts ### [Face Cropper - Tool to Automatically Crop Faces from Photos](https://civitai.com/articles/2147/face-cropper-tool-to-automatically-crop-the-faces-from-photos) **Python script for automatically detecting and cropping faces from source images.** Essential tool for preparing training datasets. Uses DLIB face detection to automatically crop faces at various resolutions (512x512, 768x768, 1024x1024) with multiple crop variations, significantly speeding up dataset preparation. 🔗 [Read Article on CivitAI](https://civitai.com/articles/2147/face-cropper-tool-to-automatically-crop-the-faces-from-photos) ### [Script to Browse Creator's Gallery Locally](https://civitai.com/articles/2328/script-to-browse-creators-gallery-locally-and-some-general-news) **Local HTML-based tool for browsing and searching through model collections.** Provides fast, responsive search with alphabetic sorting and filtering capabilities. Useful for managing large collections of models and keeping track of uploads. 🔗 [Read Article on CivitAI](https://civitai.com/articles/2328/script-to-browse-creators-gallery-locally-and-some-general-news) --- ## Model Comparisons & Reviews ### [The Big May of 2023 Comparison of ~25 Realistic Models](https://civitai.com/articles/755/the-big-may-of-2023-comparison-of-25-realistic-models) **Massive comparison of 25 realistic base models as foundations for LoRA/LyCORIS training.** Tested 77 LyCORIS models across 24 base models with 20 prompts each (36,960 generations total). Provides statistical analysis ranking models by their ability to work as base models for additional training. Realistic Vision 2.0 achieved top ranking with 71% quality score. 🔗 [Read Article on CivitAI](https://civitai.com/articles/755/the-big-may-of-2023-comparison-of-25-realistic-models) ### [Testing Some Photorealistic Models - Review Soon, But For Now Some Samples](https://civitai.com/articles/136/testing-some-photorealistic-models-review-soon-but-for-now-some-samples) **Early testing and samples from photorealistic model evaluations.** Shows preliminary results from testing various realistic models including ICBINP V8 and others. Part of ongoing model comparison efforts to identify the best base models for LyCORIS/LoRA usage. 🔗 [Read Article on CivitAI](https://civitai.com/articles/136/testing-some-photorealistic-models-review-soon-but-for-now-some-samples) --- ## Announcements & Resources ### [Introduction to Other Instance Tokens](https://civitai.com/articles/77/introduction-to-other-instance-tokens) **Discussion about using alternative trigger tokens beyond 'sks' for models.** Addresses the request for different instance tokens (like 'zwx') to enable multi-person image generation and face blending. Explores the pros and cons of standardized vs. unique tokens for different use cases. 🔗 [Read Article on CivitAI](https://civitai.com/articles/77/introduction-to-other-instance-tokens) ### ["The List"](https://civitai.com/articles/1733/the-list) **Public training queue and wishlist of requested models.** Google Spreadsheet showing the priority queue for upcoming model trainings, with supporter requests taking highest priority. Includes information about model availability and early access through the BuyMeACoffee page. 🔗 [Read Article on CivitAI](https://civitai.com/articles/1733/the-list) ### [New Version of My Base Model - Serenity V2 (for SD 1.5)](https://civitai.com/articles/3198/new-version-of-my-base-model-serenity-v2-for-sd-15) **Announcement and detailed explanation of Serenity V2 base model.** Describes the complete creation process including merging 45 models and fine-tuning with 45,000+ images including celebrity datasets and high-quality photography. Explains the merge strategy, fine-tuning approach, and model philosophy. 🔗 [Read Article on CivitAI](https://civitai.com/articles/3198/new-version-of-my-base-model-serenity-v2-for-sd-15) --- ## Quick Links - **CivitAI Profile:** [http://civitai.com/user/malcolmrey](http://civitai.com/user/malcolmrey) - **Reddit:** [http://reddit.com/r/malcolmrey](http://reddit.com/r/malcolmrey) - **Hugging Face:** [https://b-61j.pages.devm/malcolmrey](https://huggingface.co/malcolmrey) - **Support/Coffee Page:** [https://buymeacoffee.com/malcolmrey](https://buymeacoffee.com/malcolmrey) --- *All articles listed above are hosted on CivitAI with detailed instructions, sample images, and downloadable resources.*