Gabi C. Duncombe

Designer @ Adobe (Generative AI + Video)
Previously @ Instagram (Reels) & Microsoft (HoloLens)
(opinions are my own!)

[Gabi C. Duncombe @gcduncombe]

I recently found a pic on my phone of my newborn daughter holding my finger.

It made me think - everyone loves to joke about how "AI can't do hands", but I didn't think that was true anymore.

So I challenged myself to create an AI film using ONLY shots of hands.

[Gabi C. Duncombe @gcduncombe]

will only be accepting my @midjourney medical scans as expressive oil painting or watercolor --raw --chaos 65. studio ghibli if i'm feeling fragile.

[Gabi C. Duncombe @gcduncombe]

@clarkcharlie03 This is so fun

[Gabi C. Duncombe @gcduncombe]

Most product design advice is written for productivity software. Creative tools play by different rules.

https://t.co/yQDDz7Vyqn

[Gabi C. Duncombe @gcduncombe]

Here's how it came together, including the part where I quit.

I started it in Firefly Boards with text-to-image ages ago (about 6 months ago, which is a lifetime ago in the land of GenAI), back when motion-tracked text meant doing every shot in After Effects.

I generated all the video clips in Veo based on the initial images - clean, with no prompts composited in. But I didn't have time to motion-track all of it in After Effects (that's not my area of expertise), and I shelved the whole thing.

I came back to the project when GPT Image 2.0 landed with text rendering that was good enough to bake the prompts straight into the first frame of each shot.

Then I used Kling image-to-video and prompted the motion so the text tracks above each person's head. No manual tracking at all! It's not perfect, but it got the job done.

I cut it in Premiere, and voiced it myself (including some AI tooling to transform my voice into the second character). Added music from Adobe Stock.

I got some support on the sound design by the brilliant Laurent Delforge, one of the filmmakers we work with at Adobe, done entirely with our Adobe audio models and internal prototypes we've been building.

Added in a quick color pass in Premiere with the new color tools, plus some FilmImpact effects and transitions, and here we are. Enjoy!

[Gabi C. Duncombe @gcduncombe]

I prompt video models all day as a designer on Adobe Firefly's video team. Now I see the prompts everywhere. So I made a short film about it with AI 👇

[Gabi C. Duncombe @gcduncombe]

Ugh, another brainstorm on the calendar today ⛈️

(made with Gemini Omni)

[Gabi C. Duncombe @gcduncombe]

Creative work has always been fluid, but software has been rigid. The interface is the same shape for everyone. You memorize where features live and bend yourself to fit. I think that's changing, and malleable software is where things are heading:

https://t.co/0cl5PznTo0

[Gabi C. Duncombe @gcduncombe]

They said AI would give me 3x output. They did not specify the form that would take! I guess this is my life now.

(horrifying experiment made with Gemini Omni 🙂🙂🙂)

[Gabi C. Duncombe @gcduncombe]

Brought in the GenAI cleanup crew today to get my keyboard looking spiffy again 🧹

(made with Gemini Omni)

[Gabi C. Duncombe @gcduncombe]

My keyboard is always so dirty, and I think I finally caught the culprit on camera 😄

Using AI to blend between the real and generated worlds has been such a joy lately!

Google Omni helped me visualize this little critter that has previously only lived in my imagination.

[Gabi C. Duncombe @gcduncombe]

Beautiful (and informative) loading state!

[kvick @kvickart]

It's fascinating watching the generation visualization on @reve's interface.

It looks like its actually progressively editing as it steps through the diffusion process. You can see it identifies bounding boxes for objects as they appear and moves them around if they are incorrect before finishing the image generation

[Gabi C. Duncombe @gcduncombe]

Too much screen time for me. Need to go touch grass.

(made with Gemini Omni)

[Gabi C. Duncombe @gcduncombe]

A good video edit goes unnoticed, but a bad edit really jumps out at you!

Use your video editing tools responsibly🎞️

(made with Gemini Omni)

[Gabi C. Duncombe @gcduncombe]

Bring your ideas to life with @AdobeFirefly!

[Gabi C. Duncombe @gcduncombe]

@alexanderchen weirdly seems to be working now! maybe an intermittent bug

[Gabi C. Duncombe @gcduncombe]

@alexanderchen Hmm strange! Yea I tried a few different things but on my end it seems consistently blocked

[Gabi C. Duncombe @gcduncombe]

@GoogleAI I love this! It’s so fun when the physical and digital worlds come together

[Google AI @GoogleAI]

We wanted to see if we could take simple, physical materials (like cardboard and markers) and use AI to bring them to life. What was the result? A short film starring a bunch of TPUs getting ready for the big stage at Google I/O 2026!

Working with director Laurie Rowan and Nexus Studios, we kept human artistry at the center of the film by blending puppetry and 3D animation with our models to do the following ↓

Nano Banana: Generated beautifully stylized first frames from the raw puppet footage and basic 3D animations.

@GoogleAIStudio: Built a custom tool inside the platform to test these frames at scale, ensuring pixel-perfect consistency

Gemini Omni & experimental @GoogleDeepMind Models: Merged the base animation and stylized frames to elevate the final piece to a cinematic level.

Our AI pipelines were specifically designed to protect the crafty details that give these films their heart, like the tiny human imperfections of puppetry, or the nuance an animator can build into an expression.

[Gabi C. Duncombe @gcduncombe]

@alexanderchen ok you inspired me https://t.co/o4B6OHep46

[Gabi C. Duncombe @gcduncombe]

that feeling when you take off your glasses and you realize you actually can't see at all without them 🤓

thanks for the nightmares (and the fun genAI model) gemini omni!