Visual system + performance in financial services
Podemos Progresar
Product and performance for financial services.
card-marca-podemos-progresarThe process
I joined the product team and followed the build stage by stage, along the roadmap. It’s a financial product for a low-income segment, so the one constraint that ruled everything was trust: faces and places people recognize, and language with zero friction. This audience is wary — and they have good reason to be.
That’s where Chava and Liz were born: two promoters who walk members through different moments of the communication. Since they’re recurring characters, continuity stopped being a detail — same features, same brand clothes, piece after piece. I solved it with character sheets inside a workflow I built, and they turned out to be essential for keeping the video consistent.
caso-podemos-progresar-proceso-1The scenes have dialogue, so the care went into the intonation and the real slang these women use. Everyday and close, never pitying them or talking down to them. It’s one of the most human brands I’ve worked with, and it taught me a kind of advertising you rarely see: nothing fancy, just true.
Where I learned the most was building the Podemos Progresar site with vibe coding. I then took that into landing pages tailored to each marketing territory, with the character holding from start to finish: from the ad to the landing page, one single system. All of them hooked up to Meta and WhatsApp, with server-side validation and protected credentials.
caso-podemos-progresar-proceso-2Making the images
All the casting and locations were generated with AI. For this audience, financial ads that look too polished feel suspicious, so the art direction rested on three decisions.
Tight framing, direct eye contact and calm micro-expressions — that’s where trust lives. Lived-in spaces — uneven natural light, worn textures, rooms that get used — where the dream is the stability of your own small business or a kitchen that’s actually cooked in. And a precise, dignified portrayal of Mexican women: this audience has a merciless radar, and one slip of plastic skin or a blank stare sets off the scam alarm.
The consistency and the organic texture come from a controlled workflow: bases in Midjourney, composition control with Nano Banana 2 and Pro, and skin and asymmetry recovery with Magnific on selected pieces.
caso-podemos-progresar-proceso-3Instructional design
The challenge: teach financial processes and a brand-new app to women between 40 and 60, many with little tech experience, in a setting where a mistake means money anxiety. Manuals and static screenshots only add friction.
2D and 3D animation became the whole teaching architecture. Grounded in cognitive load theory, every complex flow breaks down into micro-interactions, taken one at a time. The system takes the fear out of the complicated parts while keeping the grown-up tone these users expect.
caso-podemos-progresar-proceso-4Node-based video
The enemy of AI video is inconsistency: a face that changes between shots wipes out this audience’s trust in a second. So the workflow was anchored to master sheets.
Before a single frame gets animated, I lock in character sheets and location sheets — static and highly detailed. That’s where I dial in the exact warmth of a face and how imperfect a space should look. Those assets enter the node workflow as a strict starting point for video, so the motion models inherit the lighting, textures and features already defined.
The result is continuity across cuts: the same woman, in the same home, moving through different steps of the financial process in different videos. On top of that workflow we tested agentic methods via MCP and API, building a pipeline that can produce at scale.
caso-podemos-progresar-proceso-5
