Guest
Stefano Corazza
Head of AI Research, Canva
Follow Stefano Corazza
We’ll email you each new moment, one email a day at most. Plus the best of every big interview, Wednesday and Saturday.
Check your inbox
We sent a link to your address. Click it to finish signing up. Not there in a minute? Look in spam.
or
It’s free.
Real Alpha has cut 8 moments of Stefano Corazza (Head of AI Research, Canva) talking, from an interview on The TWIML AI Podcast published Nov 10, 2026.
What Stefano Corazza said
AI Design Still Feels Like a Slot Machine
8 moments · 32 min
- 5:09
Canva's data makes smaller design models more economical
“you can fine tune or train a model that is two orders of magnitude fewer parameters if you have the data and you can beat the bigger model.”
at 5:53 - 16:34
Detailed scene descriptions help smaller models create better designs
“we underestimated how detailed of a scene description you can give to a diffusion model”
at 16:47 - 39:35
Owning the model stack creates more room to cut inference costs
“you can only make it really efficient if you own the full stack”
at 41:10 - 10:58
Layered designs let people move elements before regenerating pixels
“And so it's a lot easier for a designer to just move that element to the right and then regenerate.”
at 13:34 - 22:51
Design editing needs benchmarks beyond single-shot generation
“And you have benchmarks for image editing, which is great, but design editing is different than just image editing.”
at 23:33 - 27:04
Style mismatches can be fixed without retraining from scratch
“the cartoon images generated with craft initially were a little bit too realistic and too 3D looking.”
at 28:37 - 31:14
Brush revises designs through tools and visual feedback
“the faster you make that iterative loop, the more effective everything is”
at 35:15 - 35:48
A simpler agent harness lets the model use familiar code paths
“trying to leverage how good these LLMs are at writing code and at the same time simplifying the harness in the most basic possible way”
at 37:08
- 5:09