It got extremely popular very quickly. com Oct 17, 2023 · 画像の準備が完了したら、Stable Diffusion WebUIの「Train Tools」のタブをクリックし、「Create Project」をクリックして作成するLoraのファイル名を設定します。 次いで、「Create Version」をクリックしてバージョンを入力します。ここは初期のv1のままで大丈夫です。 Finetuning Stable Diffusion 1. Oct 17, 2023 · To make your own Stable Diffusion model, you need to collect a large amount of data for further processes. (with < 300 lines of codes!) (Open in Colab) Build a Diffusion model (with UNet + cross attention) and train it to generate MNIST images based on the "text prompt". Preprocess images tab. Jan 22, 2023 · How to use ChatGPT. 📄️ Training Status. 25/run. classUNet_conditional(UNet): See full list on stable-diffusion-art. Step 2: Train a new checkpoint model with Dreambooth. Today, we are excited to show the results of our own training run: under $50k to train Stable Diffusion 2 base1 from scratch in 7. 1. pipe = StableDiffusionPipeline. By running Stable Diffusion locally, you can experiment with different text inputs and In this video, you will learn how to use Dreambooth to train a Stable Diffusion model based on your own art style. Method 4: LoRA. to get started. Artists, get inspired by your own art styl May 27, 2024 · Table of Contents. Powered By. But it is not the easiest software to use. A) Under the Stable Diffusion HTTP WebUI, go to the Train tab The Stable-Diffusion-v1-5 checkpoint was initialized with the weights of the Stable-Diffusion-v1-2 checkpoint and subsequently fine-tuned on 595k steps at resolution 512x512 on "laion-aesthetics v2 5+" and 10% dropping of the text-conditioning to improve classifier-free guidance sampling. Once your images are captioned, your settings are input and tweaked, now comes the time for the final step. Project folder. You will learn how to train your own model, how to use Control Net, how to us Now, iirc stable diffusion uses clip embeddings, which themselves are based on gpt-2/3. The model was pretrained on 256x256 images and then finetuned on 512x512 images. Results now. Here is a step-by-step guide on how to train a Stable Diffusion model: Collect and preprocess your data: This involves gathering relevant data and cleaning it to remove any errors or inconsistencies. Train a Dreambooth Model with Custom Images. Stable Diffusion is a latent text-to-image diffusion model. Large-scale diffusion models have achieved state-of-the-art results on text-to-image synthesis (T2I) tasks. Live access to 100s of Hosted Stable Diffusion Models. 1-768. This notebook can be run with a free Colab account. In your stable-diffusion-webui folder, create a sub-folder called hypernetworks. The super resolution component of the model (which upsamples the output images from 64 x 64 up to 1024 x 1024) is also fine-tuned, using the subject’s images exclusively. A few short months later, Simo Ryu has created a new image generation model that applies a technique called LoRA to Stable Diffusion. The name must be unique enough so that the textual inversion process will not confuse your personal embedding with something else. Here I will be using the revAnimated model. ← Marigold Computer Vision Create a dataset for training →. Train Model. Veți putea să experimentați cu diferite prompturi text și să vedeți rezultatele în 2 days ago · For fine-tuning, you will be using the Pokémon BLIP captions with English and Chinese dataset on the base model runwayml/stable-diffusion-v1-5 (the official Stable Diffusion v1. 📄️ API Overview. If you put in a word it has not seen before, it will be broken up into 2 or more sub-words until it knows what it is. unsqueeze(-1) t = self. Night Cafe Studio. Software. float16 May 16, 2024 · 20% bonus on first deposit. Press “+ New Chat” button on the left panel to start a new conversation. It acts as a bridge between Stable Diffusion and users, making the powerful model accessible, versatile, and adaptable to various needs. The go-to Stable Diffusion image generator for creative professionals. Since the neural network is nothing more than a mathematical model that most likely completes all the pixels in the image, it is also possible to make editing changes by giving the image to Stable UnCLIP 2. You can adjust hyperparameters to suit your specific use case, but you can start with the following Linux shell commands. Stable Diffusion is a powerful tool with many potential applications, such as art, design, and entertainment. Use it with the stablediffusion repository: download the 768-v-ema. Create beautiful art using stable diffusion ONLINE for free. The text-to-image fine-tuning script is experimental. It's good for creating fantasy, anime and semi-realistic images. Thanks to the passionate community, most new features come to this free Stable Diffusion GUI first. Stable Diffusion 3 combines a diffusion transformer architecture and flow matching. It's designed for designers, artists, and creatives who need quick and easy image creation. ตัวอย่างเครื่องมือ Captioning. Dreambooth - Quickly customize the model by fine-tuning it. 📄️ Dreambooth Training (V2) Train a Dreambooth Model with Custom Images (V2) 📄️ Dreambooth Training. In simpler terms, parts of the neural network are sandwiched by layers that take in a "thing" that is a math remix of the prompt. 5 model). py script shows how to fine-tune the stable diffusion model on your own dataset. org YouTube channel. I just released a video course about Stable Diffusion on the freeCodeCamp. from_pretrained(model_path, safety_checker=None, torch_dtype=torch. Download the LoRA model that you want by simply clicking the download button on the page. Everydream is a powerful tool that enables you to create custom datasets, preprocess them, and train Stable Diffusion models with personalized concepts. Installing the IP-adapter plus face model. 3. Get Started. Navigate to the Stable Diffusion page on Replicate. realbenny-t1 for 1 token and realbenny-t2 for 2 tokens embeddings. New stable diffusion finetune ( Stable unCLIP 2. Apr 17, 2024 · Step 1: Model Fine-Tuning. Give it a name - this name is also what you will use in your prompts, e. Text Prompts To Videos. It is a very simple and elegant solution. For example: a photo of zwx {SDD_CLASS}. Start creating on Stable Diffusion immediately. No setup required. Stable Diffusion: How to train LoRA for Style | LoRA Part 1 is a video tutorial by Kasucast that teaches you how to train a LoRA model for style. This provides a general-purpose fine-tuning codebase for Stable Diffusion models , allowing you to tweak various parameters and settings for your training, such as batch size, learning rate Text-to-Image with Stable Diffusion. model_path = WEIGHTS_DIR # If you want to use previously trained model saved in gdrive, replace this with the full path of model in gdrive. Access Stable Diffusion Online: Visit the Stable Diffusion Online website and click on the "Get started for free" button. Get May 16, 2024 · Once you’ve uploaded your image to the img2img tab we need to select a checkpoint and make a few changes to the settings. If you’re training with larger batch sizes or want to train faster, it’s Mar 30, 2023 · Step 2: Create a Hypernetworks Sub-Folder. Documentation is lacking. Step 2: Review the training settings. from diffusers import StableDiffusionPipeline, DDIMScheduler. Then I started reading tips and tricks, joined several Discord servers, and then went full hands-on to train and fine-tuning my own models. This is due to the base model's lack of sufficient data for these specifics, leading to mistakes and odd details when working with product images. Oct 25, 2022 · Training approach. Generate images with Stable Diffusion in a few simple steps. Fine-tuning with or without EMA produced similar results. Step 3. Be as detailed or specific as you'd like. Unlike the other two, it is completely free to use. (SGD) or Adam, to train your diffusion model efficiently. No code. With a domain-specific dataset in place, now the model can be customised. 01/image for Stable Diffusion 1. This tab is the one that will let you run Stable Diffusion in your browser. Step1 : ติดตั้งเครื่องมือ kohya_ss. Faster examples with accelerated inference. Resumed for another 140k steps on 768x768 images. " Step 2. Switch between documentation themes. In this tutorial, Kasucast walks you through the process of training a LoRA model using the Dreambooth Dec 26, 2023 · Here's how to code your own python notebook in Colab to generate AI images for FREE, without getting disconnected. Blog post about Stable Diffusion: In-detail blog post explaining Stable Diffusion. Now, we do a deep dive into the technical details behind this speedup, demonstrating how we were able to replicate the Stable Diffusion 2 base model in just 6. Experience unmatched precision and control. One last thing you need to do before training your model is telling the Kohya GUI where the folders you created in the first step are located on your hard drive. A paid account allows you to use a faster V100 GPU, which speeds up the training. With Git on your computer, use it copy across the setup files for Stable Diffusion webUI. Train a Lora Model with Custom Images. Stable Diffusion is cool! Build Stable Diffusion “from Scratch”. Open your command prompt and navigate to the stable-diffusion-webui folder using the following command: cd path / to / stable - diffusion - webui. 75/run, and SDXL at $2. This guide is designed to assist you Try Stable Diffusion v1. The authors trained models for a variety of tasks, including Inpainting. 🗺 Explore conditional generation and guidance. 📻 Fine-tune existing diffusion models on new datasets. To generate images, change the parameters and run the cell. Here's the official paper. The words it knows are called tokens, which are represented as numbers. pos_encoding(t, self. Feb 18, 2022 · Step 3 – Copy Stable Diffusion webUI from GitHub. Feb 22, 2024 · The Stable Diffusion 3 suite of models currently ranges from 800M to 8B parameters. yaml, conda activate ldm and pip install -e . One of the first implementations used it because it was a rare token Dec 30, 2023 · Free. We provide a reference script for sampling, but there also exists a diffusers integration, which we expect to see more active community development. I find it's better able to parse longer, more nuanced instructions and get more details right. This approach aims to align with our core values and democratize access, providing users with a variety of options for scalability and quality to best meet their creative needs. 5 LoRA Google Colab provides an accessible solution to train a Stable Diffusion v1. com. $15 100% FREE AI ART Generator - No Signup, No Upgrades, No CC reqd. As good as DALL-E (especially the new DALL-E 3) and MidJourney are, Stable Diffusion probably ranks among the best AI image generators. 🏋️‍♂️ Train your own diffusion models from scratch. Despite their ability to generate high-quality yet creative images, we observe that attribution-binding and compositional capabilities are still Collaborate on models, datasets and Spaces. 5 checkpoint model. It operates as an extension of the Stable Diffusion Web-UI and does not require setting up a training environment. Feb 4, 2024 · TrainTrainとは？ TrainTrainは、Stable Diffusion Web UI上で簡単にLoRAが作成できる拡張機能です。 hako-mikanさんという方が24年1月にリリースしたばかりの拡張機能で、パスやパラーメータを指定するだけで簡単にLoRAが作成できます。 Nov 19, 2023 · Stable Diffusion belongs to the same class of powerful AI text-to-image models as DALL-E 2 and DALL-E 3 from OpenAI and Imagen from Google Brain. LoRA is a type of diffusion model that can be used to generate images with a specific style. Step 1: Generate training images with ReActor. Select an accurate collection of data to get as close results to the desired ones as possible. ckpt here. LORA is a fantastic and pretty recent way of training a subject using your own images for stable diffusion. We’re on a journey to advance and democratize artificial intelligence through open source and open science. Playing with Stable Diffusion and inspecting the internal architecture of the models. Step 2 : เตรียม Data Set ภาพตัวอย่าง. Training on your own dataset can be beneficial to get better tokens and hence better images for your domain. Type a text prompt, add some keyword modifiers, then click "Create. Train a Stable Diffuson v1. Dreambooth Finetunning API Overview. It’s easy to overfit and run into issues like catastrophic forgetting. Let words modulate diffusion – Conditional Diffusion, Cross Attention. Include zwx {SDD_CLASS} in your prompts. ckpt) with an additional 55k steps on the same dataset (with punsafe=0. We would like to show you a description here but the site won’t allow us. Jan 17, 2024 · This notebook can only train a Stable Diffusion v1. Apr 20, 2023 · The Replicate GUI for running Stable Diffusion in the browser Step 1: Find the Stable Diffusion Model Page on Replicate. By default, you will be on the "demo" tab. It's very cheap to train a Stable Diffusion model on GCP or AWS. Mine will be called gollum. 5 LoRA model without requiring your own GPU hardware. In this project, I focused on providing a good codebase to easily fine-tune or train from scratch the Inpainting architecture for a target dataset. This stable-diffusion-2-1 model is fine-tuned from stable-diffusion-2 ( 768-v-ema. View tutorial. Whether you're looking to visualize t = t. May 13, 2024 · Step 4: Train Your LoRA Model. 500. For free is only if you make it on your own GPU, stable diffusion extensions for training. ChatGPT is aware of the history of your current conversation. Stable Diffusion is a latent text-to-image diffusion model capable of generating photo-realistic images given any text input, cultivates autonomous freedom to produce incredible imagery, empowers billions of people to create stunning art within seconds. Define key training hyperparametres including batch size, learning rate, and number of epochs. Step 5: Setup the Web-UI. Use the most powerful Stable Diffusion UI in under 90 seconds. 5 starts at $0. Stable Diffusion Interactive Notebook 📓 🤖. Method 3: Dreambooth. This makes EveryDream 2 a flexible and effective choice for seamless Stable Diffusion training. Inside your subject folder, create yet another subfolder and call it output. Once you have your images collected together, go into the JupyterLab of Stable Diffusion and create a folder with a relevant name of your choosing under the /workspace/ folder. 98. Add any model you want. Sep 22, 2022 · Wondering how to generate NSFW images in Stable Diffusion? We will show you, so you don't need to worry about filters or censorship. Preloaded on all machines. 5 for Free. まず、Stable Diffusion Web UI 上で画像の前処理を行ないます。Train タブの配下の Preprocess Images を選択します。 ここで画像のサイズを統一し、キャプションの作成を行ないます。 Source directory: 元画像を配置するディレクトリ。1で集めた画像のディレクトリを指定。 May 5, 2023 · Ecotech City, by Stable Diffusion. cd C:/mkdir stable-diffusioncd stable-diffusion. Note: Stable Diffusion v1 is a general text-to-image diffusion Feb 16, 2023 · Click the Start button and type "miniconda3" into the Start Menu search bar, then click "Open" or hit Enter. Google Colab is free to use normally, but Dreambooth training requires 24GB of VRAM (the Nov 2, 2022 · Step 1 - Create a new Embedding. Type a question in the input box at the bottom to start a conversation. The default value for SDD_CLASS is person. With Stable Diffusion, you can generate realistic and detailed images with high fidelity to the text input. Stable Diffusion is an open-source image generation AI model, trained with billions of images found on the internet. Once we've identified the desired LoRA model, we need to download and install it to our Stable Diffusion setup. It accelerates the training of regular LoRA, iLECO (instant-LECO), which speeds up the learning of LECO (removing or emphasizing a model's concept), and differential Stable Diffusion is one of the largest Open Source projects in recent years, and the neural network capable of generating images is "only" 4 or 5 gb heavy. Jan 16, 2024 · Stable Diffusion—at least through Clipdrop and DreamStudio—is simpler to use, and can make great AI-generated images from relatively complex prompts. Say goodbye to expensive VRAM requirements and he Stablematic is the fastest way to run Stable Diffusion and any machine learning model you want with a friendly web interface using the best hardware. There's no need to use the sks word to train Dreambooth. 5, and $0. To start using ChatGPT, go to chat. This will open up the image generation interface. the Stable Diffusion algorithhm usually takes less than a minute to run. Using a few images from the user as input for a subject, the AI model is fine-tuned such that it learns to bind a unique identifier with that specific subject. 8 days. This builds on the inherent promise of technology: to The Stable-Diffusion-v1-4 checkpoint was initialized with the weights of the Stable-Diffusion-v1-2 checkpoint and subsequently fine-tuned on 225k steps at resolution 512x512 on "laion-aesthetics v2 5+" and 10% dropping of the text-conditioning to improve classifier-free guidance sampling. Feb 28, 2024 · Training a Stable Diffusion v1. Generate AI image for free. 45 days using the MosaicML platform. Generating 1024x1024 images costs from $0. According to the Replicate website: Generate. Apr 28, 2023 · In our previous blog post, we showed how we used the MosaicML platform, Streaming datasets, and the Composer library to train a Stable Diffusion model from scratch for less than $50,000. My advice: If you are new, start training 1. Example by Olivio Sarikas. Wait for the files to be created. You have multiple options for running Dreambooth. No code required to generate your image! Step 1. Stable Diffusion Online is a free Artificial Intelligence image generator that efficiently creates high-quality images from simple text prompts. The next step is to install the tools required to run stable diffusion; this step can take approximately 10 minutes. 🧨 Learn how to generate images and audio with the popular 🤗 Diffusers library. Use it with 🧨 diffusers. Apr 26, 2023 · A few months ago we showed how the MosaicML platform makes it simple—and cheap—to train a large-scale diffusion model from scratch. It utilizes the Stable Diffusion Version 2 inference code from Stability-AI and the DreamBooth training code from Hugging 🌟 Master Stable Diffusion XL Training on Kaggle for Free! 🌟 Welcome to this comprehensive tutorial where I'll be guiding you through the exciting world of Feb 18, 2024 · Stable Diffusion WebUI (AUTOMATIC1111 or A1111 for short) is the de facto GUI for advanced users. All these amazing models share a principled belief to bring creativity to every corner of the world, regardless of income or talent level. As a comparison, my total budget at GCP is now at $14, although I've been playing with it a lot (including figuring out how to deploy it in the first place). g. That being said the results are more pixelated. time_dim) return self. Describe your image: In the text prompt field provided, describe the image you want to generate using natural language. Overview. Installing LoRA Models. A widgets-based interactive notebook for Google Colab that lets users generate AI images from prompts (Text2Image) using Stable Diffusion (by Stability AI, Runway & CompVis). Started with the basics, running the base model on HuggingFace, testing different prompts. Tons of other There are a lot of factors that influence this but, in general, the average price for using Stable Diffusion online will range between $5 and $15 a month. It was initially trained by people from CompVis at Ludwig Maximilian University of Munich and released on August 2022. This weights here are intended to be used with the 🧨 Training and Deploying a Custom Stable Diffusion v2 Model. 1 ), and then fine-tuned for another 155k extra steps with punsafe=0. Stable Diffusion is an open-source latent diffusion model that was trained on billions of images to generate images given any prompt. display import display. Train an SDXL LoRA model if you are interested in the SDXL Model. 2. It is also recommended to collect the most relevant data for your task to get better results. Pretrained model name. Midjourney, though, gives you the tools to reshape your images. Step 3 : เตรียม Text Description ให้แต่ละรูป. First of all you want to select your Stable Diffusion checkpoint, also known as a model. Jul 19, 2023 · Dreambooth allows you to take any subject (person, pet, object) and put it in a Stable Diffusion model. (Open in Colab) Build your own Stable Diffusion UNet model from scratch in a notebook. Method 5: ControlNet IP-adapter face. ckpt) and trained for 150k steps using a v-objective on the same dataset. Trusted by 1,000,000+ users worldwide. Prioritizing versatility with a focus on image and caption pairs, it diverges from Dreambooth by recommending ground truth data, eliminating the need for regularization images. Jan 3, 2023 · DreamBooth is a brand new approach to the “personalization” of a text-to-image diffusion model like Stable Diffusion. Quality, sampling speed and diversity are best controlled via the scale, ddim_steps and ddim_eta arguments. 5, it is faster, and you'll learn more things in less time. 5 LoRA. Similar to DreamBooth, LoRA lets you train Stable Diffusion using just a few images, and it generates new output images with those objects or Text-to-image models like Stable Diffusion are conditioned to generate images given a text prompt. jpg files in a folder your_folder. Freemium. To read this content, become a member of this site. The train_text_to_image. 02/image for SDXL. Stable Diffusion is a latent diffusion model conditioned on the (non-pooled) text embeddings of a CLIP ViT-L/14 text encoder. This notebook aims to be an alternative to WebUIs while offering a simple and lightweight GUI for anyone to get started Feb 22, 2023 · Stable Diffusion. Most service websites give you a limited number of image generations for free. Stable diffusion makes it simple for people to create AI art with just text inputs. Nov 7, 2022 · Using techniques like 8-bit Adam, fp16 training or gradient accumulation, it is possible to train on 16 GB GPUs like the ones provided by Google Colab or Kaggle. I’ve been playing around with Stable Diffusion for some weeks now. Civitas also let's you train loras and 5000buzz cost 5$ as well which gives you like 10 loras. Dec 9, 2022 · Training-Free Structured Diffusion Guidance for Compositional Text-to-Image Synthesis. Those are the steps to follow to make this work: install the repo with conda env create -f environment. Replicate. We'll use the Diffusers library from Huggi Training Lora for custom objects can be challenging, as Stable Diffusion may not accurately capture product details, including unique shapes, cuts, and symbols. In the hypernetworks folder, create another folder for you subject and name it accordingly. This is a tool for training LoRA for Stable Diffusion. Stable Diffusion v1 refers to a specific configuration of the model architecture that uses a downsampling-factor 8 autoencoder with an 860M UNet and CLIP ViT-L/14 text encoder for the diffusion model. 📄️ Lora Training. În acest notebook, veți învăța cum să utilizați modelul de difuzie stabilă, un model avansat de generare de imagini din text, dezvoltat de CompVis, Stability AI și LAION. openai. Train Model Endpoints. Initialize your model: This involves setting initial values for the model's parameters. This model allows for image variations and mixing operations as described in Hierarchical Text-Conditional Image Generation with CLIP Latents, and, thanks to its modularity, can be combined with other models such as KARLO. แก้ This will save each sample individually as well as a grid of size n_iter x n_samples at the specified output location (default: outputs/txt2img-samples). The subject’s images are fitted alongside images from the subject’s class, which are first generated using the same Stable Diffusion model. Training a model can be taxing on your hardware, but if you enable gradient_checkpointing and mixed_precision, it is possible to train a model on a single 24GB GPU. Nov 28, 2022 · In this free course, you will: 👩‍🎓 Study the theory behind diffusion models. Elevate your creativity today. After that, most sites will either charge by the minute or a recurring monthly fee. Aug 14, 2023 · Learn how to use Stable Diffusion to create art and images in this full course. Jan 4, 2024 · The CLIP model Stable Diffusion automatically converts the prompt into tokens, a numerical representation of words it knows. Figure 1: Imagining mycelium couture. FlashAttention: XFormers flash attention can optimize your model even further with more speed and memory improvements. Use it with the stablediffusion repository: download the v2-1_768-ema-pruned. unet_forwad(x, t) The conditional model is almost identical but adds the encoding of the class label into the timestep by passing the label through an Embedding layer. What is Dreambooth? Dec 22, 2022 · Step 2: Pre-Processing Your Images. Step 1: Collect training images. This article will introduce you to the course and give important setup and reading links for the course. Put all of your training images in this folder. The extensive list of features it offers can be intimidating. Aug 14, 2023 · Lynn Zheng. Principle of Diffusion models (sampling, learning) Diffusion for Images – UNet architecture. This stable-diffusion-2 model is resumed from stable-diffusion-2-base ( 512-base-ema. For example, if you set SDD_CLASS to dog then replace zwx {SDD_CLASS} with zwx dog. Feb 17, 2024 · This trainer excels in fine-tuning models for different scales. Copy and paste the code block below into the Miniconda3 window, then press Enter. Apr 17, 2023 · รวมบทความ Stable Diffusion. Some people have reported more success with 256x256 training (which is at least faster), because apparently stable diffusion was maybe originally trained at that resolution before being upscaled. 1. Understanding prompts – Word as vectors, CLIP. from IPython. Simply put, if you want to isolate the part of it Feb 7, 2023 · It was a way to train Stable Diffusion on your own objects or styles. This course focuses on teaching you how to use . Diffusion in latent space – AutoEncoderKL. Aug 16, 2023 · Tips for using ReActor. If you trained a different class, update the prompts accordingly. put your . We recommend to explore different hyperparameters to get the best results on your dataset. Step 3: Using the model. Stable Diffusion for Inpainting without prompt conditioning. Jan 30, 2024 · Running Stable Diffusion Locally. Start by initialising a pretrained Stable Diffusion model from Hugging Face Hub. This tutorial walks through how to use the trainML platform to personalize a stable diffusion version 2 model on a subject using DreamBooth and generate new images. Not Found. Overview Install. No setup. A training notebook, obtainable through site membership or purchase, coupled with AUTOMATIC1111 Stable Diffusion WebUI, forms the core of the software you’ll need. Create a folder in the root of any drive (e. ChatGPT page. Create. 1, Hugging Face) at 768x768 resolution, based on SD2. C STAGE1: Autoencoder. Prepare to spend $5-10 of your own money to fully set up the training environment and to train a model. We're going to create a folder named "stable-diffusion" using the command line. Jun 27, 2023 · In this article, we will explore how to train a diffusion model effectively, providing a roadmap for stable diffusion developers. Train your model: This involves feeding your preprocessed Google Colab este o platformă online care vă permite să executați cod Python și să creați notebook-uri colaborative. We'll be using one of the most popular methods: JoePenna's Google Colab. While there are many advanced knobs, bells, and whistles — you can ignore the complexity and make things easy on yourself by thinking of it as a simple tool that does one thing. Input pictures can be real or AI- generated. These embeddings are encoded and fed into the attention layers of the u-net. General info on Stable Diffusion - Info on other tasks that are powered by Stable Online. Using the IP-adapter plus face model. Best for fine-tuning the generated image with additional settings like resolution, aspect ratio, and color palette. Create a talking avatar from a single image & audio voice file. qk om zm mf oj aq op ys qp qf