Upscale Conf 2025. Sometimes sequels really are good

After how much I enjoyed last year's Upscale Conf, I couldn't miss this year's major generative AI event organised by the Freepik team.

So here I am again to share my impressions of the event held on 4 and 5 November 2025, focusing on what caught my attention most and how the sector has evolved in a year.

And it certainly hasn't been a minor evolution.

Will you join me for a tour of the event?

The first day

On Tuesday I focused on the talks on the main stage and skipped the workshops, which were running simultaneously in other rooms at Málaga's Sohrlin venue.

To begin with, what I liked most was the “Beta Version” talk by Isabelita Virtual, an artist I hadn't heard of before.

She presented a lot of work that resonated with me. For example, this “Unseeing”, which felt like it could have come straight from an episode of Black Mirror.:

I also found the talk by Marten Kuipers very interesting; he explained how they apply AI to real client cases.

Here are a couple of statements I couldn't agree with more:

And this other one, which we sometimes forget:

Another curious contribution came from Tey Bannerman.

Well, he was quite a character. Even more so knowing that he's ex-McKinsey (if you've worked with them, you'll know what I mean).

Using three stories, with their corresponding AI-generated images and videos, his talk focused on preparing us for a future in which decision-makers will be AIs, and our designs (and proposals, I would add) will have to appeal to them.

A different talk, to say the least.

Unlike the more traditional ones by Joaquín Cuenca, CEO of Freepik, and the panel discussion for the Chroma Awards, the “Olympic Games” of AI-made games, films and music. This was the most boring part of the day.

I couldn't attend the talk by PJ Accetturo and was left wanting more, so I'll take a look this weekend.

Lastly, I want to mention that I took the opportunity to say hello to Javier Pastor, a Xataka journalist I've been reading for more than 15 years. It was a surprise to find him there and exchange a few words.

The second day

Here I was looking for less inspiration and more “skin in the game”: creators showing their processes.

And I certainly found it.

First, with the talk by Laura Pin, who explained her complete workflow, from receiving the client brief to delivering the final piece.

Here it is:

1. Define concept / idea
2. Define keywords
3. Structure your formula
4. Check hierarchies
5. Implement “Prompt Layering”
6. Use parameters with purpose
7. Use the editor for fine-tuning
8. Replace elements (product placement)
9. Enhance details
10. Upscale and file type
11. Add/remove and modify in Photoshop
12. Retouch
13. Color and light corrections
14. Exporting with Color Profiles

The artist carried out the first seven steps in Midjourney.

The next one with Nano Banana or Flux Context.

Steps 9 and 10 with Magnific AI.

Steps 11 and 12 are done in Photoshop.

Steps 13 and 14 in Lightroom, to give it a more realistic, analogue look rather than such a digital one.

Amazing how much the world of photo retouching and composition has changed since I worked in it, and how greatly AI has contributed to that change.

Seriously, incredible:

To obtain this final piece:

But of course, this masterclass focused on images.

So, in the next workshop, Elena Berjillos from Freepik showed us how to create a complete little social media campaign for a brand like Adidas. Starting with how to build a prompt:

Yes, the campaign was fictional.

Yes, the videos weren't the same quality as the still images (probably because they weren't upscaled), but the concept and how to do it—within Freepik, of course—were explained very well.

And when she recreated an Apple ad, first changing its protagonist and then converting it from horizontal to vertical, I was once again convinced that companies that don't make video ads simply don't want to, not that they lack resources.

That wraps up the highlights, but below I've included videos of all the talks from both days.

Not the workshops, though—they haven't uploaded those.

And rightly so: that way you'll be encouraged to come next year.

Watch the talks online

Here are the talks. The hard stuff, if you're into generative AI. Very much aimed at content creators, I should warn you.

Day 1:

Day 2:

PS1: Coca-Cola's Christmas advert

On the very days of the conference, Coca-Cola released its Christmas advert, once again generated with generative AI.

Here it is:

And, as you can see, it's simply an evolution of last year's advert—the first one the company made using AI:

Watching one after the other gives you a pretty good idea of where the quality ceiling for AI-generated video was last year and where it is now, much closer to major Pixar-style 3D animated productions.

Without quite reaching them. Yet.

PS2: La rama de oro

And to round off this week of AI, here's a podcast with some connection to all this, as it features Javi López, founder of Magnific AI, a startup focused on AI image upscaling that Freepik bought last year for a tidy sum.

In Joan Tubau's Kapital podcast, Javi, Joan and Héctor García—whom I hadn't had the pleasure of knowing—discuss what success is, how to achieve it and the path they followed.

Their outlook closely matches mine, which I've shared here before. So I can only recommend listening to it:

If you enjoyed this article, you can subscribe to receive new posts by email using the form below the FAQs.

Or follow me on my YouTube channel, for more practical content.

Do you have questions about Upscale Conf 2025? I answer them here

1. What is Upscale Conf?

It is Freepik's annual event focused on generative AI for creators: design, video, creative workflows and industry trends.

2. When and where was the 2025 edition held?

On 4 and 5 November 2025 at the Sohrlin venue in Málaga.

3. What was the standout talk on the first day?

Isabelita Virtual's “Beta Version” talk. It combines digital art with reflections on perception, featuring pieces such as Unseeing that recall the tone of Black Mirror..

4. What insights did Marten Kuipers provide?

He presented real cases where AI already delivers tangible value to clients. His central message: AI isn't hype; it works in practice. And we often forget the basics: solve real problems before chasing “the latest thing”.

5. What did Tey Bannerman propose in his talk?

A future scenario in which AIs will also be decision-makers, meaning designs and proposals will have to convince them. A provocative vision of creativity mediated by models.

6. Were there any less relevant talks?

Joaquín Cuenca's talk and the Chroma Awards panel were more conventional and offered less that was new.

7. What did the second day add compared with the first?

Less conceptual inspiration and more skin in the game: complete processes, real workflows and practical demonstrations by professionals already working with AI every day.

8. What stood out about Laura Pin's workshop?

She showed her complete workflow, from the client brief to final delivery. A clear example of how AI has transformed retouching, compositing and creative production.

9. What did Elena Berjillos teach in her session?

How to produce a mini social media campaign (inspired by Adidas) using Freepik's tools. She also showed how to adapt videos and ads—including a remake of an Apple ad—with AI and format changes.

10. Can the full talks be watched?

Yes. Freepik published the talks on YouTube:

• Day 1: https://www.youtube.com/watch?v=emTMDz9_AEA
• Day 2: https://www.youtube.com/watch?v=7xxlnCzgXaM
  The workshops are not available.

11. How does the article relate to Coca-Cola's Christmas advert?

The 2025 advert, also generated with AI, clearly shows how synthetic video has evolved since the previous year. Much closer to the standard of Pixar-style 3D animation, though it still doesn't quite reach that level.

12. Why do you mention the podcast “La rama de oro”?

Because the episode features Javi López, founder of Magnific AI, which was acquired by Freepik. They discuss their views on success, creativity and technology, in line with the article's underlying theme.

13. Will there be more content about the conference or generative AI?

Yes. The article invites readers to subscribe or follow the YouTube channel for more practical content related to generative AI.