The AI Shift

The Vault is back and updated to feature even more AI visual generation workflows.
(and it's still continues to be 100% FREE)

It now features over 250+ example images to help you understand every idea given within the workflows.

And we also have the fully customizable copy-pastable prompt blocks for you to interact with the concepts, try them out with your own reference images, and test out the outputs with what you have at your hand.

Prompts section is expanded to feature even more ready made prompts for each frequent need.

Try it out, and let me know what other workflows / prompts you'd like to see next on The Vault!

(you can access it from the sidebar or the apps tab of the community for FREE)

Surprised by the voice generation offered at Magnific (formerly Freepik) by the ElevenLabs v3 model. It has A LOT of voices to choose from (200+ different voices), and once you select one, you can also select for tonality, which is a great part to tinker with to get your outputs the way you want them. (they also tell you what languages the voices are trained on, so you don't get pronunciation problems)

Also, it is QUICK. Like, I was seriously surprised with the speed; it took 5 seconds for a 15 second voice and 15 seconds or something for a minute long voice.

And yes, you can do 2 speakers at the same time, so your conversational (or podcast like) generations feel much much more natural than if you were to generate separately and stitch together

Cheapest way to use Seedance 2.0 right now (literally $4 for 16 videos at 720p and at 15 seconds, which is how much a single 15s video costs from api, essentially 16x cheap)

go to mitte.ai
sign up

go to plans
select monthly
select basic (their lowest paid tier)
its says the tier is $16
but at the Stripe checkout add promo code = HANASUPPORT26
this discounts any tier $12
so you only pay $4 for the account, and they're running another promotion that gives double credits

you get a $12 discount + double credits, literally pay $4 for 16 videos of Seedance 2.0 standart quality (not even fast) for the price of 1. If you wanted to do 5 second videos that's even 3 times more.

Beware if you're using claude code, they're bringing in a new model soon so they transfered a lot of compute power towards testing it (probably).

What you currently are using will perform a lot worse than the usual. Double check its work and read through its changes, be fast and clear with your feedback on why the thing its doing is wrong. Otherwise you'll get this "dumber" version to roam free in your work.

The Visual Director AI is LIVE

Some of you already know about this from the shorts, but I know a lot of you probably missed it or haven't checked in a while, so here we go.

Visual Director AI is the custom AI I built that acts as your creative director for image and video generation. Not a basic prompt writer but more like a creative partner that actually thinks for you, pushes back on bad ideas, and handles the whole pipeline from "I want to make this thing" to production-ready prompts.

If you've been struggling with prompting, getting inconsistent results, fighting generation problems, or just hitting walls with image gen and video gen = this is literally what I built it for. [Plus, you can teach it your brand identity or specific styles and stuff like that using its style cards, so you always remain faithful to the styles you want to see]

Here's how it works: you get 200 credits a month with the subscription. You can choose to spend your credits in either quick-answer or deep-thinking mode. Those 200 credits will carry you over for most months. But if you're deep in a project and need more runway, credit packs are also available. Buy more messages whenever you need them. No waiting, no limits on your creative flow.

I made this open to use for anyone who saw it in the videos, but if you're hearing about this for the first time... now you know about it too. Just see the "Products" tab above the posts feed.

Tomorrow I'm dropping a full video walking through the product. Every feature, what it can do, how it thinks, the whole thing. If you want to catch that when it goes live, hit the bell on the channel so you don't miss it.

And if you want to see what it can do before you commit to anything, scroll through the shorts on my channel. You'll see what this thing outputs and how easily it gets there. (I'll link a few examples as a comment here)

I'm also offering a 1-week free trial for those who want to try it out before deciding.

Nano Banana 2 is honestly the greatest thing to have happened in a while.

It showed us an image model can get BETTER while getting CHEAPER.

Turn all your workflows with Nano Banana Pro (NBP) into Nano Banana 2 (NB2) and cut on those costs, time wait, while increasing quality.

I've been testing for a few hours but I'm still yet to see where NBP still wins, if you come across something let me know!

Jewelry size issues in AI generations. Too many of you are commenting or messaging me about it. Send me some examples, common issues you deal with, where you've given up... I'll build out something to fix this once and for all.

Comment what workflow you need. I'll build a few for vids.

You know if you were there, but the most important find of yestarday was seeing Freepik Spaces implement latest AI models to their workflow tool. Now we can intelligent prompt outputs customized per image inside the workflows.

Too much fun awaits.

Updated the Workflow Vault with consistent character generation prompts. Feel free to follow the video and use the prompts to generate your characters.

It has the "NEW" badge to help you find it easier.

Let me know what you're planning with it!

I just dropped a full breakdown of my character consistency workflow, from reference photos to multi-angle grid to final scene generation.

The key: don't describe your character from scratch. Feed 4 reference photos into a synthesis prompt that blends features into one original person, then generate a multi-angle consistency sheet (front, side, 45-degree, medium shot).

Every AI model I tested, Nano Banana Pro, Flux 2 Pro, and Seedream, could pull the character reliably from that single grid image. (Tho seedream stumbles sometimes)

What brand character would you build first with this workflow?
(Workflow Vault will get updated with a "Consistent Character" page soon so you'll have access to the prompts there)