AI model for Instagram

[ofmailo]

I want to create my first AI model but i need help and recommendations for the software to use.

Which are best ways to generate an AI model and then be able to create different images of the same model with dif clothes, dif backgrounds etc.

Which are best softwares to then turn the images to videos for reels

All content needs to be able to be made in NSFW as well as SFW

 

[Elchavo]

I want to create my first AI model but i need help and recommendations for the software to use.

Which are best ways to generate an AI model and then be able to create different images of the same model with dif clothes, dif backgrounds etc.

Which are best softwares to then turn the images to videos for reels

All content needs to be able to be made in NSFW as well as SFW

Let me give you some information. Check your private message box.

 

[ofmailo]

bump

 

[BM AGENCY]

I think you should hire a freelancer.

 

[duke1921]

Let me give you some information. Check your private message box.

Can you share it publicly in the forum here?

 

[Manual Growth]

Can you share it publicly in the forum here?

+1

 

[proxygo]

the days of a real woman are disappearing lol

 

[Nikolaos]

the days of a real woman are disappearing lol

That's definitely true!

 

[plainen]

I want to create my first AI model but i need help and recommendations for the software to use.

Which are best ways to generate an AI model and then be able to create different images of the same model with dif clothes, dif backgrounds etc.

Which are best softwares to then turn the images to videos for reels

All content needs to be able to be made in NSFW as well as SFW

You can research using search engines for this, there are recent AI app for making custom pictures and you could choose from the result

 

[Ashk881]

I want to create my first AI model but i need help and recommendations for the software to use.

Which are best ways to generate an AI model and then be able to create different images of the same model with dif clothes, dif backgrounds etc.

Which are best softwares to then turn the images to videos for reels

All content needs to be able to be made in NSFW as well as SFW

I have been researching this topic for awhile now because I am writing an ebook about it.

So here's how you do it:
- First you need to choose a base image generation model. Stable Diffusion is the most popular. But there are more options now, Flux is also very good.

- Then you need to choose a frontend/gui for stable Diffusion because doing everything by cli/manually is going to be tiring. Automatic111 was the first and most popular webui, these days ComfyUI has also become quite powerful.

- after setting up a gui with your model, you need a finetuned model file. These are called LoRAs. These are like your base model but much smaller and tune your model for certain things. You can find several LoRAs on civitai.

- for consistent characters you will need what is called an IpAdapter. For comfyui there is this:
https://github.com/cubiq/ComfyUI_IPAdapter_plus

For consistent faces you first generate what is called a *face map*, it's basically set of images of a face in different angles. You then train an IpAdapter Lora on this image.

Similarly there are ControlNets for Pose, clothing, interiors and even lighting.

It's definitely out of scope to give step by step instructions in this comment, but if I ever complete that ebook, you know where to look for if you want something very detailed about this topic

 

[Repackage7618]

Which are best ways to generate an AI model and then be able to create different images of the same model with dif clothes, dif backgrounds etc.

ComfyUI

Which are best softwares to then turn the images to videos for reels

Nothig really, everything is either censored, low qualiry or extremly pricey. Not really worth doing it, AI videos are doing rather poorly on any SM

I have been researching this topic for awhile now because I am writing an ebook about it.

So here's how you do it:
- First you need to choose a base image generation model. Stable Diffusion is the most popular. But there are more options now, Flux is also very good.

- Then you need to choose a frontend/gui for stable Diffusion because doing everything by cli/manually is going to be tiring. Automatic111 was the first and most popular webui, these days ComfyUI has also become quite powerful.

- after setting up a gui with your model, you need a finetuned model file. These are called LoRAs. These are like your base model but much smaller and tune your model for certain things. You can find several LoRAs on civitai.

- for consistent characters you will need what is called an IpAdapter. For comfyui there is this:
https://github.com/cubiq/ComfyUI_IPAdapter_plus

For consistent faces you first generate what is called a *face map*, it's basically set of images of a face in different angles. You then train an IpAdapter Lora on this image.

Similarly there are ControlNets for Pose, clothing, interiors and even lighting.

It's definitely out of scope to give step by step instructions in this comment, but if I ever complete that ebook, you know where to look for if you want something very detailed about this topic

Pretty much this, if you are starting from 0, expect 3-6 month to learn all of it, after that you can pretty much do anything

 

[iAi]

Merhaba. Yaklaşık 1 yıl önce oluşturduğum, 90'a yakın gönderi ve 88 bin takipçiye sahip Ai Influencer hesabımı sizin için satabilirim. İlgileniyorsanız özel mesaj gönderebilirsiniz.

 

[imccafey]

There's good free ones available for generation now, but for the best ones or at least the ones I've seen being used on IG or Twitter, it's usually the paid ones which have been custom trained. Stable Diffusion is the common one, I'd say. You can usually get a hang of it within an afternoon. Use tools like ControlNet with it for consistency in features while making changes to certain elements of what you're generating. Particularly on the NSFW side, you might need to experiment with other Stable Diff-based AIs for variety. It's a massive plus if you have decent PS skills to be able to make convincing edits.

 

[ZenMarkets]

posting for reference

 

[sdmins]

I've had good luck with Fooocus for image generation. Took a bit to get the right loras and models, but now that it's set up I can generate a handful of decent photos quickly. Their Faceswap is pretty good for keeping a consistent image.

 

[By_ReFLeX]

To produce content with AI, your spending on hardware alone will be around 20k USD. If we add AI account fees, this figure will be 30k USD

 

[shijin]

Check out StyleGAN2 for creating your models, If you need those images turned into high-quality videos, use Final Cut Pro or Adobe Rush to animate and export for Instagram reels.

 

[Repackage7618]

To produce content with AI, your spending on hardware alone will be around 20k USD. If we add AI account fees, this figure will be 30k USD

You can run local SDXL models on hardware as low as GTX 1080/RTX 2060 which cost around $150. There are also many free online providers but I can't really recommend them

 

[MR.Spuf]

I have been researching this topic for awhile now because I am writing an ebook about it.

So here's how you do it:
- First you need to choose a base image generation model. Stable Diffusion is the most popular. But there are more options now, Flux is also very good.

- Then you need to choose a frontend/gui for stable Diffusion because doing everything by cli/manually is going to be tiring. Automatic111 was the first and most popular webui, these days ComfyUI has also become quite powerful.

- after setting up a gui with your model, you need a finetuned model file. These are called LoRAs. These are like your base model but much smaller and tune your model for certain things. You can find several LoRAs on civitai.

- for consistent characters you will need what is called an IpAdapter. For comfyui there is this:
https://github.com/cubiq/ComfyUI_IPAdapter_plus

For consistent faces you first generate what is called a *face map*, it's basically set of images of a face in different angles. You then train an IpAdapter Lora on this image.

Similarly there are ControlNets for Pose, clothing, interiors and even lighting.

It's definitely out of scope to give step by step instructions in this comment, but if I ever complete that ebook, you know where to look for if you want something very detailed about this topic

Great miniguide, appreciate it!

 

[recdown]

I've had good luck with Fooocus for image generation. Took a bit to get the right loras and models, but now that it's set up I can generate a handful of decent photos quickly. Their Faceswap is pretty good for keeping a consistent image.

I've been trying Loras in Fooocus for a while now, but I'm not getting the results I'm looking for. Can you give me some Loras tips and recommendations for Fooocus? I'm looking for focused backgrounds, as in a normal photo, and that the skin does not come out as if it were plastic.