Future of E-commerce?! Virtual clothing try-on agent
HTML-код
- Опубликовано: 3 июн 2024
- I built an agent system which will autonomously iterate & generate img of AI model wearing certain cloth and produce millions+ social posts
Free access to run any comfyUI workflow, hands-restoration model, upscaler & more on Replicate: replicate.fyi/ai-jason
🔗 Links
- Follow me on twitter: / jasonzhou1993
- Join my AI email list: crafters.ai/
- My discord: / discord
- Get the source code: forms.gle/yYaE4WvzY7qHW3j16
- Install ComfyUI: github.com/comfyanonymous/Com...
- Deploy your comfyUI workflow: replicate.fyi/jz-any-comfyui-...
- Hand-restoration model: replicate.fyi/jz-hand-restora...
- Ultimate-portrait-upscale: replicate.fyi/jz-ultimate-por...
- Install insightface: • ComfyUI InsightFace Wi...
- IP Adapter (FACEID) Models: huggingface.co/h94/IP-Adapter...
- IP Adapter Models: huggingface.co/h94/IP-Adapter...
- ControlNet (SD1.5) Models: huggingface.co/lllyasviel/Con...
- ControlNet (SDXL) Models: huggingface.co/lllyasviel/sd_...
- Checkpoint Models: civitai.com/
- More in-depth comfy UI tutorial by Data Leveling: • ComfyUI IPAdapter (SDX...
⏱️ Timestamps
0:00 Intro
0:43 Use case
2:01 How does img generation works
5:37 ComfyUI Basic
07:21 Method 1: Integrate cloth into existing image
09:38 Method 2: Finetune model & IPAdapter
10:55 Install comfyUI manager, custom nodes & models
14:30 Step 1.1: Generate AI model face
19:51 Step 1.2: Embed cloth img
21:43 Step 1.3: Hand restoration & Upscale
24:00 Step 1.4: Deploy image generation pipeline
24:43 Overview of how to build self-improving agent system
26:12 Step 2.1 Setup agent tools to call AI models
29:30 Step 2.2 Setup multi agents system
34:02 Demo
👋🏻 About Me
My name is Jason Zhou, a product designer who shares interesting AI experiments & products. Email me if you need help building AI apps! ask@ai-jason.com
#gpt5 #stablediffusion #openaidalle #autogen #gpt4turbo #gpt4 #ai #artificialintelligence #tutorial #stepbystep #openai #llm #chatgpt #largelanguagemodels #largelanguagemodel #bestaiagent #chatgpt #agentgpt #agent #autogen #autogpt #openai #aiimages 
Наука
What kind of comfyUI image generation pipeline do you want me to build next?
Interior design pipeline for empty rooms
Im actually keen to see if we can take this a step further, can we create a 1 min (or close) video after we create the model with the style/clothe of choice? that would be rad if possible
Therapist
Build a pipeline that can dress a person with real clothes (image reference) via IP Adapter
Keep the image same but chage the sunglasses 😎 using a set or collection of designer sunglasses
Man's building entire venture backed companies in 30 min videos
yea
recently i saw a startup booth with the same idea
@@Abubakr844 Do you remember the name of that startup ?
This guy just made me realize the work I did on my ad campaigns is nothing special. Thank you for humbling me, keep up the good work!
This guy gives real life practical projects. Thanks as always!
Hey Jason thanks for the mention! Really appreciate it and love your post-processing towards the workflow process, awesome work!
I watched many videos which shows things in bits and pieces. But in this video you put all the bits and pieces together and present it with a real world example where the output is really real world standard. Absolutely amazing. This one video is really better than many ComfyUI playlist put together. I learnt a lot from this video. 🎉
This is absolutely incredible. I am extremely excited to be able to use this project. What a game changer. Thank you so much Rockstar!!
DAYUMMM MAN, this was literally a very insightful and never-ending movie masterpiece. I was amazed about the tools you've showcased and for sure I will try to give it a go and learn to use, but at a certain point I was like: WAAAT? It's still going? DAYYYUUUM MAN. I'm impressed about each video you're dropping. You've done excellent job in the past videos, but this one is going beyond. You're my inspiration and never cease to surprise me. I will try my best to be as amazing as you Jason. Keep it up man! See you in the next one, for sure :)
The only channel with real world examples
The hand fix model you shown is pretty amazing, gonna try it out!
This is awesome AI Jason. Thanks a lot!
Definitely one of the best complex workflow. Thanks @AI Jason
Wow, bravo. Nice job incorporating agents into the workflow
Another kick ass video Jason! This is the future :)
I didnt try img generation much before as i always got the impression that it is not easy to control, but man, the comfyUI looks freaking powerful!!
Jason, you are so amazing!
Man, you deserve a prize!! Awesome piece of content, lovve it
I don't have the technical capacity to follow a tutorial, even though I have access to the same Linksm tool, I teach courses on the subject for non-technical people, teach me how to create my own model for Ecommerce, this is gold, for the adult niche there are several ways to monetize this.
love this video. Thank you. I'm going to be using it for my men's fashion site.
This is really dope man
This is so freaking cool! It would be so cool if AI progress enough for us to produce videos that demonstrate the functionality of the product too, just by giving it a picture of the product and the list of functionalities. That would be the future of Internet influencers and ads.
16:54, i dont have this prepare image nodes. How can i get that?
Amazing content, you never disappoint. I have recently started dropshipping, and I am trying to follow every guide I come by, and now I am having a store with over 300 sales per week. I just used products from winnerzila for my store, and I had conversions from first days, at least 8-10. Of course, I am not yet making millions but I hope to be there one day.
Hi...i want to start dropshipping..but a little confused regarding it....aree u open to help?can u please share ur whtsapp here?
With great respect to you that you extend the explanation in an exaggerated way.
Mr. Jason, you are just amazing. I hope I can one day work with you on my project together. Keep going. :) Many thanks.
Thank you for sharing. That's awesome.
awesome workflow
Good work ,you are so amazing!
Dude you are amazing!!! An absolute legend, a total beast, a god!
One thing might has your attention, here is a breaking update on the ipadapter plus happened at 2024/03/23, the old ipadapter nodes used in this workflows won't work. FYA. Thanks for doing this demo, awesome.
Thank you for this insightful video
Wow mind blown. Amazing
many great ideas, thx
Best AI channel out there.
How about one with voice command for ecommerce?
As always awesome detailed video with perfection buddy :-)) in future Amazon and other similar websites will be using these projects to create model and put different clothes on them for showcasing instead of hiring models plus if any fork of this project with reduced complexity is developed in which people can put there faces too and see how they might actually look like easily while wearing particular pattern of clothes then many would like this feature I guess
Definitely! The question is, however, whether it is also possible to incorporate product-specific details such as logos and fonts. Has anyone already had experience with this?
Sick! Now I want to play with comfyUI
can you try putting virtual clothing on a 3d model, because i guess it will be more relevant to the company by showing perfect body fit
@aijasonz Awesome work yet again AI Jason! Would love to know what other youtubers learn from as well as other sources you find to be helpful in growing your AI skills.
Good work ! Very Impressive
This channel is giving away free start up ideas and I absolutely love it!! ❤
This is so cool
Thank you without even trying it yet.
Great content!
Amazing content in 30 mins
Interesting idea with using an agent for that! Just in time for me :) I've trained a LoRA on my subject but the generating and upscaling take too much time. I only wonder how much does it cost with gpt4-v and running the model on Replicate? They both were quite expensive the last time I used them
cool video, thanks
Thanks for sharing! But still, how to use your own face image (not generated) and lets say, copy pose or enviroment, to make it more controlled?
Can comfyUI be uploaded to some server like vercel and use this image generation (on auto-pilot) or some kind of make trigger?
Bro, I was waiting for something like this for my shop! This will help me a lot, thank you a million times!
Awesome video! Just one question, why comfy vs a1111?
Hello - I am following your tutorial. However at 17:00, the node "Prepare Image for InsightFace" is not available. I have installed all the custom nodes as shown by you. Can you please guide me where I am going wrong.
Thanks!
Hey Jason, great video.
Can you make a similar guide with automatic1111?
That would be wonderful.
Hi, possible to have the final workflow comfyUI ? Thanks for the video
thanks 🙏
Freaking amazing
You explained img generation process so clearly, thanks a lot! Is the video generation similar process but adopt to each frame?
would love to know
Damn this comfy AI looks really awesome. Long live Open Source!!
Hey Jason, why not create a SAAS for this use case? Would sell damn well
I can't find the node Prepare Image For InsightFace, and Apply IpAdapter FaceID, Can you please let me know what is the possible reason ?
@AIJasonZ can you add the code to description? Thank you for your knowledge sharing! 🙏
does this work on other accessories likes necklces, rings, and holding products by hand like perfumes, ointment ....
@AI Jason, I just did a fresh install of ComfyUI. I dont see 'prepare image for insightface' node. How do I add that?
same here, did you find out?
cubiq commented last week
this is not needed anymore as the image is automatically optimized
@@MikeDaUnicorn cubiq commented last week
this is not needed anymore as the image is automatically optimized
hey can you make a detailed video on Ipadaptor installation guide for comfyui
@aijasonz what computer would you recommend for doing all these ML tasks on? Anything affordable :)
Hi Jason, I'm following your video, but at minute 16:56, when searching for 'Prepare Image for InsightFace,' only 'IPAdapter InsightFace Loader' appears. Am I missing something? Thank you so much for this tutorial!
same here, did you find a solution?
@@MikeDaUnicorn +1
Samesies
@@MikeDaUnicorn I am looking for a solution for this as well. Let me know if you have found it mike
@@sahir1059 have you found the solution for iy sahir?
OOTD = Outfit of the day
Are you and Immortal Cultivator? Why are you so great?
Nyc video .. can you pls also tell us the cost involved running these processes
Thanks
Thank you Dave!
You’re an absolute boss, I hope you’re making $1 million a year
Is there a way I can start learning this from scratch? I am currently not good in basic dsa
Im not getting the prepare image for insightface , dont know where im going wrong
It's fascinating to see how AI-generated influencers are reshaping the landscape of social media and the fashion industry! Leveraging AI for generating social media posts not only enhances customer engagement but also opens up new avenues for visualizing products and reaching diverse audiences. 🤖✨
omg models should be shaking right now xD
What I am supposed to do about the bin files?! everything in my ipadapter folder is safetensors not bin, you did not explain that
goodness wow
I was searching for this for months. I wanted to try out cloths how will they look on me before buying. So, this is perfect for me. But, why I can't I access the replicate model? It is redirecting me to the explore page.
How I can use this ? Just uploading clothes and model only , not entire process???
is it free for commercialization?
i never try comfyUI before, only Stable Diffusion
can you tell us the different between both of them?
maybe from the PC usage
i have RTX3050 6gt and 24ram with i5 intel
but when every time i tried to do train image it's take to long in stable diffusion.
can comfyUI also do train image? it's much more easier or faster in comfyUI than SD?
So stable diffusion is the model itself, where comfy UI is a platform that allow you build image generation pipeline (where you can use multiple models)
I’m using 3080 and can get it working so should be fine I think (though gonna be slow)
Sadly this won't work, because even smallest details will make a dress become a different one. Even in your demo, you chose a fairly easy to replicate design but it still resulted in a different one at the end. And that will be fraud if we deliver a different item to buyer
hello i need help i cant find Prepare Image For InsightFace on confy interface
can you make a video on how to run many of these replicate models locally for free?
Hey as a complete freshers,not able to understand what is happening in this video but i like to create same.can you provide simple roadmap for this
I was able to follow up until generating the face image, then all the IPadapter nodes and preview bridge nodes are not showing up. I'm not sure if I'm doing something wrong or its just that the nodes are broken (as some say in the comments) would love to know whats the problem here. thanks!!
Me too brow :/
Bro, Prepare image for insightface seems to be removed and all the other nodes from ipadapter got renamed. Can you provide some help regarding that? BTW great video again as always.
anbody else facing this issue? solution?
@@aviator033 not yet mate, ipadapter guys said they will push an update soon to bring everything back, we gotta wait though
@@aviator033 23:19 is the answer
any updates?
@@blackboy004 not yet
I could make a video of a model in which I can choose 2 clothes for the photo created, for example leggings and a sports bra, and this model I can use another photo as a reference for the "influencer" pose created, for example doing exercises
15:35 what do you mean? I dont get it, i have to what? Please explain to me. Add to where???
🎯 Key Takeaways for quick navigation:
00:00 *🌐 La influencia creciente de los modelos de IA en las redes sociales*
- Exploración del uso de modelos de IA como influencers que parecen personas reales y acumulan seguidores.
- Discusión sobre el valor comercial potencial y aplicaciones prácticas de estos modelos de IA en la industria de la moda.
02:04 *🖼️ Fundamentos y aplicaciones de la generación de imágenes con IA*
- Descripción detallada de cómo funcionan modelos como la difusión estable y su aplicación en la generación de imágenes hiperrealistas.
- Uso de modelos preentrenados y descripción de procesos para personalizar la generación de imágenes adaptada a necesidades específicas.
05:35 *🎨 Integración y personalización avanzadas en la generación de imágenes con IA*
- Implementación de técnicas para insertar elementos específicos en imágenes y personalización detallada usando modelos avanzados.
- Ejemplos de aplicaciones prácticas como la prueba virtual de ropa en imágenes generadas, destacando modelos específicos para la industria de la moda.
09:29 *🛠️ Desarrollo y ajuste fino de modelos de IA para generar imágenes personalizadas*
- Técnicas para el ajuste fino de modelos con conjuntos de datos específicos para mejorar la generación de imágenes con características deseadas.
- Utilización de adaptadores IP y otras herramientas para una personalización aún más detallada en la generación de imágenes.
19:51 *🧥 Integración avanzada en la generación de imágenes personalizadas*
- Proceso detallado para integrar nuevas prendas en imágenes existentes usando modelos avanzados como el IP adapter.
- Se muestra cómo ajustar los modelos para incorporar elementos específicos en las imágenes y cómo mejorar los detalles utilizando prompts específicos.
21:38 *🌍 Despliegue y optimización de flujos de trabajo de generación de imágenes en la nube*
- Descripción de cómo desplegar flujos de trabajo de IA en plataformas en línea para mejorar la accesibilidad y rendimiento.
- Explicación sobre la interconexión de varios modelos de IA para mejorar la calidad de las imágenes y cómo ajustarlos para su uso en plataformas específicas.
Made with HARPA AI
I am pretty sure that Temu have been doing this for some time
Hey how to get the source code? Is there any github repo?
That’s great but overly complicated in my view. The backend is scattered over comfy UI viz programming spaghetti that is hard to maintain and debug (what will happen to my deployed app if comfy UI service goes off? How would I recover and maintain the source code?) …and using autogen agents swarm just to refine and upscale the image is hell of overkill in terms of resources needed. Can we think of more efficient solution to the same problem? Just sharing thought, great content overall
how can i make my own can u help me with that
Can you please help ? I could not find prepImageforInsightface Node
I do not see a node called "Prepare Image For Insightface"
MAKE SOME NOISE FOR THIS MAN
Hi everyone, I'm having a problem, actually it's more of a weirdness than a problem, I did all the installations as in the video, but there is no prepare Image For InsightFace node, I can't solve it even though I did a lot of research, is there anyone who can help?
my comfyui is not generating images
I really want to get access to this as an API, i will pay