Organic Acquisition Systems for Agency Owners and Execs - Founder @ Voxen & ScalingDavids

Hi there 👋,

I'm Andrea, and here’s my story in a nutshell:

🌍 Originally from Italy, I've lived in 7 countries and learned 5 foreign languages during my early 20s! 

🚀 I worked at Meta in Sales for the EMEA markets, consulting with 500+ businesses and entrepreneurs on growth marketing and branding. 🚀

🏗️ In 2020, I founded and grew Scaling Davids, a Creative Marketing Agency focused on video ads production. 

💡 I'm building and growing Voxen, a personal branding studio that enables Agency teams and executives/founders to grow their personal brand and drive influence at scale. 📈

📊 From zero to 7 figures and a team of 20 talented individuals without any funding or previous knowledge. 
🤝 Trust-building and genuine value-based relationships have been key to getting here so far. I'm currently building a personal brand online to form relationships at scale.
💸 I'm sharing my ongoing learnings and lessons, hoping to inspire people and businesses to be better.

I’ll be posting tips and tricks about content, growth, and how to build an effective branding strategy - so hit that “Follow” button!

Peace out ✌️

Andrea Tonetti

Fino a 6 mesi fa A Casa Loro era un sitarello su Wordpress creato in fretta, che iniziava ad accogliere il traffico dei primi utenti. 

Oggi esce una storia su di noi su La Repubblica (grazie alla fantastica Martina Ciai per l'occasione), di fianco a nomi ben più "navigati". 

E' bello vedere quanto una visione forte e fresca attiri attenzioni inaspettate, e siamo solo alla fase 0-1.

Ne vedremo delle belle ;) 
Link nel primo commento!

Ieri sera A Casa Loro ha segnato la prenotazione numero 100, e prima che il jet lag accumulato mi ammazzasse ho registrato due parole a riguardo dall'aeroporto👇  

Da una idea buttata quasi per caso a 100 clienti, un team di 7 matti, un network di 10 partner tra aziende, giornali e università in pochi mesi. 
Il tutto con unit economics profittevoli dal giorno uno, no capitali esterni e offrendo esperienze che fanno bene per davvero a tutte le parti coinvolte.

Bella storia, per davvero.
Mo a lavorare, e ci si risente a 100 prenotazioni/mese ;)

KFrame Interactive sa

Art Director

LSV multimedia SRL

iconmedialab

Sergio Valsecchi

Did you know Nano Banana Pro supports this kind of syntax?
I recently found and tested a hidden feature using GPS coordinates + date + time.

Here’s the exact prompt I used:
“Make a photo at 31.7784° N, 35.2296° E on 05/04/0033 at 04:30:00 UTC.”
The output?

Absolutely insane.
The model interprets coordinates, timelines and historical context with a level of coherence that feels unreal.
At this point, prompting is officially turning into a new programming language.

#nanobanana

Flux2 or Z-Image Turbo: which one really wins?

Sometimes I’m convinced AI should just give us 36-hour days. You spend the whole week thinking the latest model drop is going to change your life, and then barely 18 hours later Z-Image-Turbo shows up: fast, sharp, surprisingly clever, and in a few cases actually outperforming Flux2.

After putting it through real tests, the difference hits immediately. Z-Image-Turbo understands prompts with almost surgical precision, fires back results at near real-time speed, and keeps everything clean and consistent without that subtle CGI sheen you still catch in some heavier open models. It’s not perfect, but the overall balance of detail, natural texture, and low artifact noise is impressive.

Flux2 is still a beast. It’s heavy, demanding, and built for people with serious hardware, but the technical punch is undeniable: advanced light handling, ultra-detailed renders that can push past the 4-megapixel mark, FP8 optimizations that squeeze VRAM usage, and a realism pipeline that can look ridiculous when it hits the sweet spot. 

The trade-off is that it eats resources and slows down your whole workflow unless you’re running a monster setup. Z-Image-Turbo plays a different game. It’s quick, light, prompt-aware, and easy to plug straight into a professional pipeline without rethinking your entire stack. And the quality is surprisingly high for something that runs this fast.

I’m not saying it rewrites the rulebook, but it definitely forced me to rethink what a “lightweight” model can do.

What do you think? Should I post a more technical breakdown or show some visuals?

#zimage #flux2

We are no longer standing in front of technology; we are inside technology.

I see it every day in my work.
Clients give me a mood, a couple of directions, a photographic intention and the AI handles the rest.
No prompt engineering, no syntax to remember: the workflow interprets, understands and builds the composition for you.

In Adding People 1.0, I distilled this idea into a workflow.
A pipeline where ComfyUI, paired with my custom nodes, analyzes the talent, the environment and the outfit, then constructs the perfect prompt for Nano Banana 2 and generates the final scene without altering the original environment.
You provide the vision, the system handles execution.

It marks an important shift: creativity is no longer a set of commands.
It’s a direct conversation with technology........... faster, more natural, and much closer to how we actually think in images.

Workflow link: https://lnkd.in/d_gfyM_r

#comfyui #nanobanana

Today a client asked me: “Sergio, what do you think AI will look like a year from now?”

My answer was pretty simple.
You’ll write a brief to an AI agent, and it will send back exactly what you asked for.
And yes… I’m building that system, and it’s almost ready.

In this post I’m sharing a few cosmetic ads that come straight from that idea.
They were generated from just a short written brief using the alpha version of my new workflow. No fancy prompt tricks. No manual retouching. You describe the intention, and the system puts together talent, product, layout and message with surprising clarity.

The workflow blends an LLM with Nano Banana or Flux.2 and hooks into Morpheus Model Management, which automatically picks the right talent for each creative direction. I’m already working on the next step: turning the same brief into full campaigns, including all the video formats needed for social ads.

What interests me isn’t only how good the images look.
It’s what this shift represents.

For years we stood in front of technology, trying to follow its pace.
Now we’re slowly stepping inside it. We shape it with our intentions.
Creativity becomes an ongoing conversation between what we imagine and what the system can generate back.

--- There’s still one big piece I’m working on: real-time metadata analysis for each market. ---

Imagine campaigns that update every single day because an agent reads, interprets and evaluates everything happening online around your product. It can correct wrong impressions, highlight new opportunities or reveal insights the company didn’t even know existed… or didn’t know how to look for.

This direction is heavily inspired by the work of Massimo Giacchino on book ANALISI AI DI MERCATO

We’re moving toward a space where briefs become conversations and ideas become actual systems.

And these cosmetic ads you see in this post are just the very first signal of what’s coming next.

#comfyui #kerastase #loreal #k18 #theordinary #ai #flux2

Today I ran a solid round of tests comparing Z-Image Turbo and Flux 2, and here are the first takeaways.

It is hard to declare a clear winner. Flux 2 surprised me with its micro-detail rendering, while Z-Image still handles human facial structure more naturally, even if it sometimes introduces a bit of noise that hides some of the finer details.

The next step is to test a few LoRAs and see how much they can push the overall consistency and quality depending on the type of project you want to build.

You can find the full comparative workflow here:
 https://lnkd.in/dMQa_ZZP

#flux2 #zimage

Took FLUX.2 for a spin.
What do I think after a proper test drive?
------------------------
If you want to experiment hands-on with FLUX.2 prompting, I built a custom GPT that analyzes any image you upload and generates a clean, ready-to-run FLUX.2 prompt. If you need it, it can also output a full JSON version following the official Black Forest guidelines.

Quick tip while you’re at it: try the dpmpp_2s_ancestral sampler or the res_x family. They tend to produce sharper micro-details and cleaner structure.

Drop a comment and I’ll share the GPT link.
------------------

FLUX.2 hits hard. The quality jump from the 1.x series is obvious, the prompt understanding is on another level thanks to the new visual backbone and the built-in multi-reference system, and the realism is honestly impressive.

On the flip side, it’s heavy. Really heavy. For 99 percent of users running 8 to 12 GB GPUs, pushing it to full power is almost impossible. The dev model runs at 32B parameters and wasn’t exactly born for budget hardware.

Nano Banana Pro or FLUX.2 dev?
Nano Banana Pro lives inside Google’s ecosystem with access to absurd amounts of training data, so it’s incredibly stable straight out of the box.

FLUX.2 dev, on the other hand, lets the community go wild: open weights, local inference, custom optimizations, LoRA support, product-level detail control, skin refinements, and all the RnD playground we love. The new open VAE also helps with finer texture clarity.

If your GPU doesn’t have 24 or 32 GB of VRAM, good news. A few hours after launch, the community already pushed GGUF-optimized builds:
https://lnkd.in/d7MmaZug

And the final showdown? Qwen, FLUX or Nano Banana, Image-Z?
Way too early to pick a winner. With the pace models are dropping, three months from now we’ll all be hyping something entirely new.

For now, I’m enjoying the ride: high fidelity, a brand-new visual intelligence core, true multi-reference support and huge potential for serious ComfyUI pipelines.

#comfyui #flux2 #nanobanana

In recent months, we’ve often discussed what model management will become in the age of artificial intelligence. Not just generated faces, but constructed identities, designed, shaped, “educated” as if they were actors born from intention rather than from a body.

But now the question expands, and becomes almost philosophical.

We live in a world moving rapidly toward personal robotics.
First they will be assistants.
Then coworkers.
Then everyday presences in our homes and public spaces.
At that point: will robots become influencers too?

It sounds amusing, but it’s not science fiction. If a robot has an appearance, a character, a visual style, and a set of recognizable behaviors, then it also has communicative value. And if it has communicative value, someone will want to hire it.

This weekend I tried a small experiment.
I took Model 626, one of the profiles from my Morpheus Model Management archive. I transformed it into a robotic figure, dressed it in an #ON outfit, photographed it in studio, and then “took it outside,” as if it were a real outdoor photoshoot.

It wasn’t the copy of a human.
It was something else. A new aesthetic, an alternative physiognomy, a body designed for a world that may not exist yet, but that we are all beginning to imagine.

And here lies the central point:
perhaps AI is not here to replicate reality.
It is here to open new communication pathways, new visual codes, new forms of presence.

In parallel, I continue working on invisible watermarking:
a cryptographic digital filigree, embedded directly into the noise of the image.
A kind of synthetic DNA that links every output to its model, its rights, and its licensing conditions. Because if these “new visual beings” become relevant, we will need systems that protect them.

I’m building a library of 1,000 identities.
I’m around 700: human, hybrid, and now robotic.
Each will have its own trio of LoRA models (WAN, FLUX, Qwen) to guarantee full coherence across any production pipeline.

Maybe it’s a useless project. 
Or maybe it’s a way to prepare for a future in which brands won’t just choose models, but aesthetic agents born in the digital realm.

And so yes, it begins to make sense to ask: What happens the day robots want to join the casting too?

👉 Instagram: https://lnkd.in/dMEfs-bZ
 Next week I will release 700 models on Patreon.

#comfyui #on

Every few weeks, the world loudly asks the same existential question:
“Can AI truly improve society?”
And every time, we philosophers of the digital era try to answer with nuance, ethics, and a hint of techno-realism.

But let’s be honest: nothing triggers a deeper reflection on the future of humanity… 😁 
 …than seeing someone upload a pizza with pineapple and asking an AI,
 “Is this authentic  Italian cuisine?”

At that precise moment, the entire Mediterranean cultural heritage collapses like a poorly drained pot of pasta.

So I decided to give AI a chance at redemption.

I built a free workflow where you upload any dish, and AI,  with the diplomatic tact of a silicon Socrates.... generates a flowchart and infographic explaining how the dish should actually be prepared.
- Qwen VL identifies the dish (hopefully without crying).
- Nano Banana 2 / Pro turns the recipe into a visual guide.
- SeedVR2 Upscaler makes it look like a Michelin-star presentation.

>>>> https://lnkd.in/dwAiVnzP

Is this the technological leap humanity needed?
Will this transform society?
Will it prevent the next person from putting cream in carbonara?
Probably not.

But at least, for once, AI does something truly valuable for collective well-being:  it protects real food from digital chaos.

And honestly, if AI can save even one pasta dish from being violated…
 …maybe there is hope for the future after all.

Stay tuned.

This is just the first of 50 workflows I’m releasing for Nano Banana users.
If AI can’t fix the world, it can at least fix dinner



#nanobanana2 #nanobanana #Italiancuisine #food

taplio

The Battle

Andrea Tonetti

Sergio Valsecchi

Followers

Posts

Engagements

Likes

Performance ScoreAndrea Tonetti's LinkedIn Performance

Performance ScoreSergio Valsecchi's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Andrea Tonetti

Sergio Valsecchi

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Andrea Tonetti's LinkedIn Performance

Performance ScoreSergio Valsecchi's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreAndrea Tonetti's LinkedIn Performance