FLUX

Recently, we've been sharing more about AI video modeling and less about graphical modeling. We haven't shared about flux for a long time, so today we're going to share about flux! Today we will continue to introduce the acceleration techniques of flux, we have also introduced a good article about flux accelerated graph acceleration techniques, there are accelerated by Lora, and torch.compile, and the most recent is TeaCache and WaveSpeed, etc. With the support of these acceleration techniques, flux can be accelerated by Lora, torch.compile, and TeaCache. With these acceleration techniques, flux can be used locally with 409....

When we disassemble the benchmark video, or do the AI video merchant list test demo, we often encounter such a thing. How to mimic the corresponding style of the benchmark video split screen and use it in your own video split screen? Here are a few ways to recommend to you from easy to difficult. AI style reference AI is a very convenient tool that can help us get inspiration for the target style. By simply uploading a reference image or describing the target style, AI is able to generate a style similar to the target video style for your reference. Its advantages are: Simple operation, no need for complex configuration. Suitable for quick conceptualization...

With Flux releasing the family bucket, Redux is an excellent tool for maintaining consistency. There is no need to worry about excessive differentiation of scenes and characters when doing multi-screen creations. Some enthusiasts have combined the PuLID (byte-developed plugin) face-swapping feature with Redux to achieve near-perfect one-click face-swapping of graph-born images. This article introduces the workflow, with nodes and models to download. PuLID uses version 0.91 model, due to PuLID plugin and the latest version (as of December 4, 2024) ComfyUI conflict, this article is attached to a modified version of Pu...

Introduction to AWPortraitCN Today we introduce a Flux LORA model focusing on the appearance and features of Chinese people: AWPortraitCN. The model is specifically optimized for the appearance and aesthetic features of Chinese people. The model is trained using several types of portrait images, including indoor and outdoor portraits, fashion and studio photos, etc. AWPortraitCN has strong generalization ability. Compared to the original version, AWPortraitCN is more detailed and realistic in skin texture. Suggested Trigger word: None, Suggested LoRA weight: 0.9-1. LIB...

Today's article introduces you to one of the newest and best virtual try-on solutions: catvton-flux, which combines the power of CATVTON (CatVTON: Concatenation Is All You Need for Virtual Try-On with Diffusion Models) with the power of Flux redraw repair models for realistic and accurate clothing transfers. The catvton-flux solution is also supported by the Ali In-Context...

Previously released Stable Diffusion 3 model, because of excessive auditing, resulting in the generation of human structure "regression", was sprayed a lot. Now an excellent flat replacement has appeared, that is, the Black Forest Labs FLUX model, the team consists of former core members of Stability AI. FLUX was released, the praise was overwhelming, was named "the strongest open source literate map model". Compared with the head of closed-source tools such as Midjourney, DALLE, FLUX's performance is still remarkable. At present, FLUX already supports ComfyUI graphic ...

The already powerful Flux has announced a batch of official tools covering commonly used drawing features, making it comparable to (or even surpassing) closed source tools. First you need to have the latest version of Flux installed in ComfyUI, see here: FLUX Testing and Installation Below is an itemized list of methods used in ComfyUI and the download paths for the models/workflows. First, Flux repainting (Fill Inpainting) Need to download the model flux1-fill-dev.safetensors: https://hug...

Midjourney and Stable Diffusion are both popular AI image generation tools that generate high-quality images based on text descriptions. Both are text-to-image generation models based on deep learning techniques, but they are each based on a different larger model. However, a model that has recently been introduced that is more powerful than the first two, generates more realistic images, and is more in tune with the real world in terms of details is FLUX! What is Flux? Flux AI is a newest text to image generation model by Black Forest Labs...

Today's article introduces a smart ID photo application: HivisionIDPhoto, a practical and systematic tool for the intelligent production of ID photos. With this tool, you will be able to produce a personal photo ID from an ordinary photo of an ordinary person in just a few seconds. You don't need to spend money to go to a photo store to make a photo ID. Specific features for HivisionIDPhoto can use a set of perfect AI model workflow, to achieve a variety of users to take pictures of the scene recognition, keying and ID photo generation. Specifically includes the following features: lightweight keying (pure offline ...

Technology media outlet The Decoder published a blog post on October 3, reporting that Black Forest Labs has newly launched Flux 1.1 Pro and introduced the BFL API in Beta form. Black Forest Labs (BFL) is a German-based startup focused on developing state-of-the-art generative AI models. Flux Introduction Flux is the "best" in the field of AI generative mapping, with excellent performance in a number of areas: generative quality...

Flux PuLID Introduction In the previous article has introduced the Flux model in the latest style consistency maintenance and migration components: Pulid-Flux. this is the byte Pulid team following the SDXL Pulid version of the precipitation on the basis of the release of the PuLID-FLUX-v0.9.0 Flux version of the style consistency components. Can provide a no-adjustment identity ID consistency and customization solutions, can be applied to the style consistency to maintain the field include: character portraits, continuous multi-image cartoon display and video and other areas as one of the core issues. ...

Today, we recommend a real-life realistic lora model trained on Flux: Flux_Xiaohongshu Real Style丨Daily Photos丨Extremely Realistic. As soon as you look at the naming of this model, you can guess that it is trained on the dataset of real daily photos taken by Xiaohongshu, and the style of photos generated by this model is extremely realistic and natural. Today, we travel to famous attractions in China to take pictures of the card as a scene to experience the effect of image generation. Note that the current Flux model can not correctly understand a variety of famous attractions in China, so the prompt word about the famous attractions in the scene description can be referred to, we do not have to be too serious ha. Model ...

AI painting, can be said to have been quite mature Closed source model Midjourney, easy to use, can produce photographic works, the effect is awesome. It is a few dozen knives per month, a little bit of money, and there is poor scalability. Character scene consistency requirements, or want to use the workflow to do exclusive painting tool applications, open source model is YYDS, such as stable diffusion series, and the recent fire Flux, media platforms have been slaughtered by the two lists. The charm of open source, lies in the scalability, controllability, but also with workflow packaged products, and the current effect is constantly approaching Mid...

Today we introduce a Flux LORA model released by author @Slayer of Thousands: F.1-Rebel-Girl-Tattoo-Girl-Tattoo-Atmosphere-Beauty, which is a Flux LORA model featuring a rebel girl tattoo. It belongs to the character Lora, adding tattoos, and can be used as a full body tattoo or partial tattoo Flux Expenditure Figure. Model base trigger word: Tattoos; Recommended LORA weight: 0.8. As the LORA model straight out of the figure character is more monotonous, I used a mixture of the LORA model with the Ink Spectre net map, the specific weight to see the next section of the node legend. F.1-Rebel...

Today, we recommend a Flux-based training lora model of Chinese style illustration:F.1 Chinese Style Illustration-Four Seasons Palace Shadow. This lora model of Chinese style illustration has a delicate drawing style, good recognition of ancient Chinese architectural (palace, etc.) elements, and can be drawn in all seasons. It is very easy to create a beautiful atmosphere of the national style illustration pictures. Model download address (the end of the article net disk address can also be obtained) LiblibAI: https://www.liblib.art/modelinfo/d0314e9e5cd04d8aa0805cdf2a9d9...

Given that many partners who just started AI painting want to play Flux, but on the one hand, the computer configuration is not enough, on the other hand, the unfamiliarity of the workflow, so that many partners are deterred. Today, I would like to share with you how to experience Flux in LiblibAI website. LiblibAI website: https://www.liblib.art/ About LiblibAI website here will not be detailed, no understanding of the partners can refer to my previous article. Stable Diffusion [Basic] [SD Getting Started ]：Using Liblib....

I don't know if you've all used BizyAir Cloud Node yet, but it's an open source plugin dedicated to ComfyUI by the Siliconflow team, and it requires the API key on the siliconflow website to use it. But since it's an API, you can use their functionality wherever you can run code. After playing with the coze platform for a long time, you may feel that the current raw model doesn't fulfill your needs, isn't it? Even if there are partners will be a variety of workflow building, but the local may not have enough graphics card resources, what should be done? Today to coze how to call F...

Today and share with you 9 kinds of characters commonly used lens keywords, I hope to help you in the creation of more perspective choices, these tips apply to Midjourney, Flux, SDXL and so on a variety of models The following I'll take Flux as an example to show the effect of various perspectives 1, full body shot Full body shot to show the whole body of the character's modeling. Such as clothing, posture or the relationship between the character and the environment. Like you stand in the city street, a friend to help you take a picture from head to toe. For example, a full body shot of a young college girl standing on the street waving her hand and mentioning...

Flux is a quality base model, but Flux still owns the common problems of other open source models in Chinese female portraits. Today we share with you a female Lora model trained on Flux base model - Chumhum FLUX | INS Filter. The Lora model for the portrait to add Ins style filters at the same time the ultimate optimization of the performance of Chinese women out of the picture, very oriental beauty, out of the picture of the beauty of the portrait of pure sweet. The model is trained with the help of WellCap using completely artificial natural language manual marking, cue word comprehension is very strong, based on Flux.Dev training...

Today, we would like to introduce a big model based on Flux training by Ink Specter team: Ink Specter Casual - F.1. This is a model that mainly solidifies the big style into daily casual style, which is completely different from a bunch of photographic models with refined texture, and pursues the texture of common photos in daily life, with a kind of very simple sense of realism, which is very suitable for self-promoting media and circle of friends contestants. The current version is V1.0, the author explains on the official website that the number of training sets used in this version is relatively small, so its local stability (such as hands) and diversity still need to be improved, so please look forward to the next version after expanding the training sets. Model ...

Joy Caption Model Introduction Today's article will introduce an excellent image inference model: Joy Caption. This Joy Caption model, developed by the author Fancy Feast, is based on Google's SigLIP model and Meta's latest Llama3.1 model. It uses the Adapter adaptation mode and is carefully trained to produce an excellent image inference description LLM model. It can output corresponding image description prompts with rich details according to the parameters set by the user. Google's...

Previously, I shared with you a lot of loar models based on Flux. As the Flux ecosystem matures, large models based on Flux have gradually appeared. Today, I will share with you a large model trained on Flux: the real-life FluxAura large model. The official website does not provide any relevant introduction to this model, but the official website sample image effect removes the portrait look of the Flux base model, and the realistic portrait of the characters is very fresh, the overall picture quality is delicate, and it is full of artistic conception and beauty. Model download address (the network disk address at the end of the article can also be obtained) LiblibAI: https://www.…

Today I would like to share with you an ancient style loar model trained based on Flux: Ancient style girl Hanfu. The official website does not provide any relevant introduction to the model, only the usage instructions of the loar model. However, as the current Flux ecosystem becomes more complete, various domestic loar models based on Flux have also appeared on the stage. It must be said that the open source ecosystem is really beautiful. Model download address (the network disk address at the end of the article can also be obtained) LiblibAI: https://www.liblib.art/modelinfo /825ccb21c5ec4…

Since the launch of Flux, the latest dark horse literary model, the wave of Flux is sweeping the world with an unstoppable momentum. Currently, the local computer environment to play Flux, the main support method is ComfyUI. However, the installation of ComfyUI tools has always been a confusing place for many newbies. In WeChat groups and private messages of friends, the environment configuration problem of ComfyUI has always been one of the most frequent problems. Currently, there are two main ways to install ComfyUI: Method 1: Use ComfyUI native installation package Method 2: Use Qiuye's Comf…