Intro
With FLUX being in full-steam-ahead mode, I figured it’s a good time to draft up an article to promote working together to get more buzz, improving model/content visibility, and maybe even making an internet friend or two along the way.
The intention of this article isn’t to be a ranting article about how users need to pay creators, how CivitAI needs to better incentivize users/creators, or how creators need to stop being greedy with early access. This is about collaboration.
While I've split this into different user categories for a better article flow, please don't think of this as 'users' and 'creators' being different people.
Users
These are tips for all users/creators on the site to get more buzz and engagement from the platform.
React to content
It seems a bit obvious, but the easiest way you can get buzz and show other users you found their content was interesting in some form is to hit those reaction buttons. Was it funny? Hit that laughing face button! 😂😂😂
Follow cool creators
Not only do you get buzz for the first 3 users you follow every day, but that user gets a boost in followers which provides at least a minor dopamine hit.
Search for subjects of interest
Using the search box at the top of the site, you can look up models/images/collections related to a subject of your interest.
If you like furry chicks with huge butts, chances are there's 20 other people creating that same subject - why not search it up, react to their content, follow them, collect their content to your "furry butts" collection, or remix some fuzz-butts for yourself?
Post to model galleries
This is MY BIG ASK: If you got a good generation from a model, please post it to the model's page!
This ensures the model creator gets a nice buzz reward (50-200 buzz depending on subscription)
Cross-posting does not work for this!!
Example: I use the RD Pixel Art Lora and get an image I really like. If I post that picture directly on the model page gallery, sinatra will receive a nice buzz boost.
If I were to post it on the FLUX page, or my own wall, sinatra will receive nothing (asides a few buzz in generation credit).
Review/comment on models
Following on the last point, if you liked a model: give it a thumbs up and/or provide feedback!
You can also give it a thumbs down, but unless the model is totally broken, I would suggest posting on the model asking questions of the creator/model before going this route.
Engage with creators
Did someone make a model you really like or pictures you enjoy, but you can't quite get that same prompt to work for you? Maybe you have a suggestion for improvement or a new model? Leave a comment on the picture/model and hopefully the creator will reply.
You can direct message them as well, but I would advise against it if you're going to ask for something out of the blue. Not everyone is responsive, and not everyone will accept a random message request out of nowhere.
But don't harass them!
Maybe you wasted buzz on a model due to the model being unoptimized, the onsite generator running into a bug, or you just prompted it incorrectly, this is not a good reason to DM someone asking for buzz or reporting content.
If there's nothing on their page indicating they're taking LORA requests, or text explicitly saying they aren't, I would not recommend DMing someone asking them to make you a model.
Create collections
If you really like those furry butts, go to your Collections Page and create a collection to store furry butts in (or in my case, I'm keeping a chronology of the Skippy the Clown Saga).
Every time you add an image/model to your collection that belongs to someone else, the creator of that content gets 2-8 buzz (depending on subscription)!
Tip
Ok, I know this one will get pushback, but the most direct way to tell someone you liked what they posted or that their model worked for you is to Pony up the penny (what is it with 10 buzz tips anyways?) and send a tip on a picture, model, or directly on their user page.
With FLUX being considerably more expensive to train for and generate on, tipping has gone down considerably. I hope to see this improve over time as FLUX is optimized to reduce in generation/training costs, but I understand this is a big reason for an increase in early access models as well.
Provide generator feedback
The last easy way to get buzz is to provide the onsite image generator feedback for how well it did creating images based on your prompting/models. Hit that thumbs up to indicate a good result, and the thumbs down (with feedback) if it screwed up what you wanted.
Creators
These are tips targeted towards creators to improve visibility and usage of created models.
Make a good title
Nobody should have to guess what your model is when they read the title! If it's a character, your title should follow the format of "[character name] - [series/movie/game/source name]"
For example, a good title would be "Vicar Amelia - Bloodborne", where a bad title would be "Bloodborne vicaramelia-v1.2flux"
Exceptions include standard known abbreviations like FLCL and MST3K.
For styles and concepts, it's a little more open, but anyone reading the title should be able to understand what it does, such as "Slavic Squatting Pose" or "Tetris Art Style".
(contributed by NanashiAnon)
Tag your models
The easiest way to make your model more visible is to ensure you've filled in the tags in the Edit Model page:
This will help ensure your models are seen by people looking up certain terms that may help them find your model.
Make a good content gallery
Generally the default training-generated photos are pretty weak.
Moreso with FLUX which defaults training images to half resolution.
Either by downloading your model post-training and creating new pictures locally, putting the model into early access, or setting the model live, you should 'Edit Post' on your Showcase to add in a few better examples to truly show off what your model can do.
From there you can Rearrange pictures putting your favorite (PG recommended for visibility) example from the model right up front so anyone casually browsing models can get a sense of what your model will do:
Write a good description
You don't need to write a novel like I do sometimes, but you should at least add a description of the model to explain what your model is (is it a character? a pose? clothing?) and how to use it (keywords, strength, other notes).
Search does look into the model's description, so you should include some relevant information in the description to help both the internal site search and web crawlers like Google.
Add in any special notes that might be useful to users, such as what might help get good results from the model, how you trained it/what materials you used, whether it works in certain styles/with certain checkpoints better than others, etc.
Add in some of your tagging information from training as well. Maybe certain words will help improve the function of your model for those using it as well (e.g., 3d, game screenshot, speech bubble, medallion...)
(with contributions from NanashiAnon)
Use keywords and strength
Nobody wants to guess how to use your model, especially with more costly models like FLUX.
Be sure to update the version info to include any tags needed for activation and update the strength if it needs more or less to be used successfully (e.g., if the model runs best at 1.4 strength, don't leave it at 1).
Include this information in the model description as well.
Make it remixable
If you made your pictures locally, you may need to add back in prompt information through the 'Edit Post' feature, especially if you used ComfyUI.
Remixability is key for usage - in general users (self included) would rather hit remix and change things up with the prompt than start from scratch.
Try to make your tagging clean and presentable as well - it's difficult to remix something that's just a mess of words and commas without even proper spacing.
You can use the word 'BREAK' as a logical end to a section to help separate parts as well - such as doing model/triggers, character specific tags, clothing, pose, environment, etc. that won't break the system flow.
Good example:
score_9, score_8_up, score_7_up, score_6_up, source_furry, BREAK
solo, rocko, 1girl, shirt, red footwear, sitting, legs spread, white panties, bra, BREAK
blurred background, living room backgroundBad example:
score_9,score_8_up,score_7_up,score_6_up,source_furry,dof,solo,blurred_background,living_room_background,winking,detailed_background,nothing,source_anime,rating_questionable,rocko,1girl,shirt,redfootwear,sitting,snout,legs_spread,white_panties,bra,what,dontquityourdayjob,booba
(with contributions from NanashiAnon)
Early Access
We all have thoughts and opinions on putting models up in early access to try and recoup training costs/make some buzz, but I would put out a caution against asking too much. The problem is there is no right or wrong answer for what's too little and too much right now.
I like to keep my models free and open because my favourite thing on here is seeing what kind of creative results people get from using my models, but I do understand people wanting to get back some of that time/buzz investment.
That's in part why I'm putting up this article to promote other ways to provide creators buzz. Please refer back to the Post to model galleries section for the best way to do this at no cost to yourself (asides generation).
CivitAI Mods / Staff
You thought you were going to get away without a callout? Think again!
Continue developing new features
Of all AI image generation services, CivitAI has grown tremendously over the last months (RIP old servers) largely in-part due to the community features and user engagement on the site. Keep on creating new features and taking in user feedback to keep improving the platform and soon enough you'll need to upgrade those servers again.
Feature a variety of content
I have my biases against what I consider to be lazy content being featured (and having that itself be featured), but it's been nice to see a variety of pictures on the frontpage lately: creative, funny, thought-provoking, animated, etc.
It's also been nice to see more variety in the users that have been featured, with models like Commodore_64 Pixel being featured instead of just power users.
Keep engaged
The engagement of the moderation team/staff on the site, Discord, etc. really separates Civit from other image generation services. While there may be disagreements in how matters are handled from time to time (Discord chat is a graveyard for sanity), it's nice to see a moderation team as engaged and quick-reacting to ongoing matters.
! The Buzz Cheatcode !
To truly optimize buzz earnings, you have to subscribe.
Sorry, there is no huge secret here, but paid subscribers get 50-300% more buzz.
I'm sorry if you were expecting a real secret here, I'll send you 10 buzz if you ask for it in the comments below.
Conclusion
What are your thoughts? Do you have any other ideas for additions to this article, or parts you disagree on? I'm not an expert on any of this, but I figured it was a good time to put an article together.
Please feel free to comment your thoughts on this article and I'd be happy to update it further, address concerns, etc.
If you need me, I'll be at the pool.