OneDrive to rule them all -> VIsionAPI ->-> Social Media + DeviantArt w/NSFW Routing & Logging -- VoidXedis CDN V2-

Greetings,
About a year and a half ago I started up a hobby, it was a bit controversial at the time but I persisted at it. Generative AI Art. I’ve trained my own fine tuned models, created my own LoRAs of myself and friends, I’ve generated over 500k images (not all are even remotely good, but I kept them all).

With having so many images I found myself struggling to post to social media to try and spread some of the magic I felt I had given life to. Countless hours were spent crafting social media posts, each site being a bit different, some with aspect ratio requirements, some with content restrictions, all requiring a relevant description/title/tags for each image.

Then I found Make…at first, I didn’t think it was going to actually work out…until something clicked in my head and APIs and OAuth2.0 started to finally make (ba-dun-tss) sense.

Through countless hours of struggling the attached Blueprint is the fruit of that labor of pain and love, the most efficient way to distribute customized posts while still logging any failures. Greatest achievement was setting up the custom Http node for DeviantArt, their API is somewhat dated, it’s been neglected and a lot of newer components of the site lack availability from the API

To use the following blue print you will need:

  • 1x OpenAI API Key ( access to vision model required)
  • 1x Onedrive webhook
  • 1x DeviantArt Account --Additional setup required on DeviantArt account, you must create an app.
  • 1x Twitter/X Dev account – Both API v2 and API v5 access required, it can be tricky but keep at it.
  • 1x Google Account
  • 1x Pinterest Account
  • 1x Facebook Group you are an Admin for
  • 1x Instagram Business Account (required for API access)
  • 1x Discord account w/server channel to send to.
  • 1x MySQL/Maria DB with a table schema matching the logging nodes.

The flow is as simple as dropping your image into your OneDrive folder, the automation does the rest. If OpenAI finds the image to be NSFW → the automation will route it to NSFW approved sites, with the required Mature tagging included.

If you find this blueprint useful or if you like all sorts of art (tho i do tend to generate on the nsfw side more frequently than not) grab a sub or a tip, I’ll never turn away a penny. :slight_smile: https://links.xedis.me

VXCDN-v2.json (320.7 KB)

Feedback appreciated and welcome. Questions please feel free to ask away!

Thanks,
VoidXedis

Disclaimer: Some platforms have restrictions on automated posting if you reach spammer levels, so don’t! I won’t be responsible for accounts being banned because of abuse, be smart.

7 Likes

Welcome to the Make community!

It looks like you have automated your massive supply chain with Make – well done and excellent job!

Thanks for showing off your masterpiece of a scenario, it looks like it must have taken you quite a while to set it up with lots of sweat and tears trial and error.

  • Could you shed some light as to the custom HTTP module for DeviantArt? What was the challenging bit about their API?

  • Does the OpenAI analyze images module cost you a lot to run? How many images do you put through this scenario per week/month?

  • After developing this scenario, did it affect your opinions on using Make? Was it the right tool for the job? Did you find it enjoyable to use?

2 Likes

Sorry for the delayed follow up, totally went to gather the details… ‘oh shiny… and off i went…away for a few days’
Thanks Make was definitely the right tool for the job. Not being a developer myself, (security analyst by trade) there’s no way I could have built out a solution to encompass all the different platforms. Sure they are all REST APIs but each one is a bit quirky in their own ways.

https://developers.deviantart.com <–written back in 2011

OpenAI Usage For January

OpenAI Total Bill for January
image

More up-close shot of the workflow

1 Like

Hello, I am New here! However, this workflow is intriguing and I’d like to attempt to upload and it and connect the various parts to make it work.
So I will embark on this journey to do so!
If I have any questions, can I send an email to you ?

Natala

1 Like

All feedback whether it be questions or suggestions for improvements are welcome, we can start a discussion thread in Q&A category might be best…as it’s likely you wont be the only one with them :slight_smile:

Thoughts @samliew ?

2 Likes

Hi @VoidXedis :wave:

Starting a discussion on this topic in Questions & Answers is definitely the right approach! Feel free to do so @Natala :slight_smile:

Hello. Just an update.
I could not get the database connected , but for my purposes I don’t need to log ant data - so I unlinked the database from each scenario. The posting to social media is excellent !
Note: I also disconnected Deviant Art since I’m not signed up there yet.
For part two, I may dare to connect Deviant Art some day in the future.
Thanks for showcasing this scenario ! This is Just what I needed as new comer to this space !

3 Likes

@Natala Glad to hear you got it working!

The MySQL DB step was just my choice for error handling, you could link it to a google sheet just as easily or totally remove the logging and just use the Ignore

Important to note that each of those was connected using the Attach Error Handler option thus they are only triggered when the action they are tied to fails, i just got a bit creative with it for the NSFW route.

I actually found I had to rework that step, I had an NSFW image slip thru and got me a mark on FB (RIP)

Once I get the flow cleaned up again and validated I’ll be sure to upload it.

Next up on my TODO since I was able to get an open source vision model that lacks any content restrictions (NSFW) running locally is to see if I can find/build a connector to my local instance… which would replace the OpenAI component, or at the least add equivalent for NSFW workflows as well.

Cheers!

1 Like

Cool! Show it off!
And Thanks again for sharing your showcase!

V3- Update!
-No SQL Error Logging
-Verified working import blueprint
-Returning reaction count in progress and incomplete but started.

vxcdn-v3.json (452.5 KB)