(h4)

H4 Face Fixer-Upper 3000Why 3000? Why the heck not, I thought it sounded betterbut I am a wee bit immature so who knows.As the name suggests this workflow will fix faces that have issues.Got a face that's deformed? Missing an Eye? Ugly?Give it a go, Load an image and run.Adjust the positive to suit your image. I'd recommend copying the prompt. Or simply using quality tags. Your callAll the bestH4(b ' . ' )b [Approved]

H4 Generate with Custom FaceThis idea (though I'm sure not original) came to me the other day. Wanna generate and then place someone elses face on a ridiculous or epic image?Now ya can in one go. Upload the image of the face you wanna swap, the more centered on the face or detailed theimage the better when it comes to faces. Then fill out the prompt and generate your epic image. This works just like standard image generation. You prompt for whatever you want using whatever model you want and generate. This will work for anime but the face recognition is designed for realistic. But I'm not going to stop you from plastering your own face on Deku >_< lolWhatever makes ya happy. EnjoyUse responsiblyH4(b ' . ' )b [Approved]\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\DISCLAIMER\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\This can be used to make literally anything with anyones face. That being said you're responsible for how this is used , I'm not. I made this tool for fun , not to replicate anything specific. Use at your own risk and understanding you're taking all the risk if any. ////////////////////////////////DISCLAIMER/////////////////////////////////

H4 Image2Video***Update***This workflow has been throwing exceptions, as well as for some reason it'll saveeach individual frame opposed to combining them into a singular gif.This is something on tensors end, the workflow works phenomenally locally.Which means that this can occasionally generate all the frames OR throwan exception. - Updated the Workflow, made some changes that were causing excessiveduplication of images to be generated without combining. They still do this on occasion, I dunno why, hopefully Tensor can sort this out soon.Also, for some reason the option for pingpong isn't showing up in the tool. As such I've left pingpong on, so it'll play forward then in reverse***Update***This ones a bit of a complex workflow, it has 4 individual sampling stepsin order to apply detail step by step. It uses animateddiff to apply the animationsteps to the original image. The end result is an animation effect.Sometimes it comes out epic just as you expected, other times it just appliesanimated effects. This is because the model you choose can't interpret the originalimage during the img2img phase of this workflow.That doesn't mean it wont work, it just means it may not turn out how you expect.Keep an open mind and just have fun with this. On that note:What else does this mean? It means that this can throw a lot of exceptions right now,it's not the workflow, it's Tensor. (No disrespect)The concept behind this is simpleChoose an image you want to animate or apply animations too(is what I should say).Enter a prompt, try to stick to the image as much as possible, it is possible for this to generatea completely random image if the prompt doesn't match. Then generate.Few things;This workflow can take some real time right now on tensor.This is a hardware restriction not a workflow one, that being said be patient.It's possible for this to throw exceptions again I say , it's not the workflow.

H4 Face Swap - ImageSimple workflow to swap faces on images.It uses the reactor node which currently on Tensor are throwing random exceptions. Simple procedure, choose an image to upload and the face you want to swap it and swap. I've attached an example of the swap (compare the cover with one of the showcases images)I've also attached in the showcase some very high resolution faces for use if you want. Ta-Da(b ' . ' )b [Approved]H4

***Update***For some unknown reason, it's generating images with serious color burn , that's if it letsyou load in a XL checkpoint. I'm getting this addressed now***Update***H4 Prompt Matrix v4.5 - Hi-Res Pass with Face Fix ((updated workflow))My workflow for SDXL, it's not to complex but it's definitely complicated looking.Main and Secondary Positive prompts are both for your image prompt.Feel free to switch up whats in them, however sticking to the theme and overall ideaof the image is generally recommended. Theres no need for a negative hooray!Cause I already put one in. If you want a specific negative prompt lemme know.This uses my workflow to generate images for SDXL, it can work with other XL like checkpoints,on the condition that they use similar conditioning.Choose a model, write a prompt and go, simple stuff :)If you like my work Star and FollowAll the best (b ' . ' )bH4

H4 Text2VideoBasic Workflow for text2video.Currently the showcase shows some rudimentary renders of 16 frames.This workflow, takes some getting used to, however once you get your head aroundit it's quite simple.The model you select will determine the style of the animation, not just it's quality sopick wisely ... or not , I'm not your mother.The prompting takes some adjusting to too.Example prompt should be present in prompt box. However ;Syntax breakdown"frame" : "prompt here","frame" : "prompt here",Example:"0" : "solo, 1girl, sitting, reading book","10" : " solo, 1girl, sitting, looking at viewer",The above has the potential to generate an image of a girl reading a book in the snow.Remember , AI generation is just random noise >_<Also, remember the damn Comma at the end of each frame line.I forget that thing all the time and it returns Code Errors.Tensor has it's limits, I've uploaded in the showcase a few of those examples.In other news, Sometimes it throws out just pure noise.This is because of Tensor OR it's because the frames aren't matched.The frame rate between the save and starting must be the same.This can also throw exceptions at us. Examples of pure noise are also in the showcase. They're the weird golden sparkly imagesStar if you likeFollow if you like :)H4

Batch Upload and Upscale with Face fixBasic workflow giving the ability to upload up to 6 images and upscale them while repairing faces.Since tensor doesn't have an option to upload our own locally generated images.Well heres the workaround :) So basic functions:Upload 6 images , you need all 6.Select model Modify settings, update prompt etc...RunObviously the model you use will determine if it works out well or not.I've upscaled Anime with a Realism Model and gotten some weird stuffBut stick to the style and Any model of the same style should produce good imagesThis uses UUSD to upscale so it doesn't require a high denoise.If you go to high you'll change the image. UUSD works like a proper samplernot just an upscale node.Thanks to Tensor for the AI tools finally working again!Now heres the thing;It requires all 6 images to work, thats a Tensor limitation not me.Also, if you like my work follow and star :)I appreciate it and it's nice to know people are enjoying my endeavors.(b '. ' )b [Approved]All the best

UPDATE: VIDEO NOW SUPPORTED TO UPLOAD! Huzzah! Every GIF in the showcase was made with this tool, so I know the workflow works.It's just a troublesome tool at the moment, It'll throw exceptions with longer videosbut should run fine now. I've added the ability to choose the format to save , I'm unsure if thatssupported but Let's see if we can render some low file size mp4s eh?Tensor is still having issues, which I suspect is hardware issue not implementation of the comfy workflow system. Which is unfortunate as Tensor.art's idea of what a good social site should be (in regards to our shared topic and hobbies) is IMHO, a good one, but they can't seem to properly implement it.H4 Video Face Swap - ReactorSimple reactor workflow, to swap faces in videos.The only limit in the video is time frame, which how long depends on the system being used.I've also found that higher resolution videos along the lines of 4k/8k/UHD slow down the generation and depending on how tensors feeling that day may fail.I can run this workflow on a 4 GB and 16 gigs of ram machine , while still quickly and accurately swapping faces. On a low mem machine the time limit for medium to low quality videos is 15 seconds, for high quality 10 seconds. On my high end Machine, 12 GB mem, 64 gigs ram, it can swap 30 seconds UHD fine. Struggles with more than 1 min. Though again this is not the workflow, this is the machines limitations.I've attached some high res faces I used to train face models.They're in the showcase. These are just some celebrities that I enjoy watching.\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\I'm not responsible for what you do with this, it's a tool designed to be fun and swap faces on gifsand the like, however it is capable of more. Just Note you use this at your own discretion and risk//////////////////////////////////////////////////////////////////////////////////////////////////////////

*** Fixed ***Basic upscale workflowNothing fancy here, just an upscale. Load the image , adjust your settings and go. Lower denoise ensures the continuity of the image , while higher can all out change the image that you want to upscale. It's advised not to go above 0.7 for Denoise during the upscale sampling. (b ' . ' )b [Approved]H4

H4 Prompt Matrix v4 - Hi-Res pass w/face fix (Real/Anime)*** Updated Workflow ***Complex txt2img workflow for high detailed images with clear proper faces.This tool works for both Anime and Real, just select your model,This is designed to run SD 1.5. It does not conform to SD3 or XL standards.Nor does it work with Flux. Works just like any other site out there,Choose your modelLoad a lora if you want oneAdjust settings and continue.This workflow typically on my local machine produces about 11/itsI can generate a 768x1280 image in under 20 seconds.Higher resolutions demand more obviously.I'm working on making it cheaper per run on Tensor.Now Keep in mind thats my local machine, so in theory it should run quickeron Tensors Hardware. However we are limited on the sites usage and queue.Upcoming Tools:PromptMatrixV5 SD1.5PromptMatrixV4 SDXLPromptMatrixv4 Flux