Idea to production: AI apps with Tomek Porozynski
Year: 2026 · ▶ Watch on YouTube
Aja Hammerly (Director DevX AI) · Tomek Porozynski (Google Developer Expert, Cloud & AI; Staff ML Engineer)
Segments (8)
- 00:00:00 · Introduction — Aja Hammerly
- Aja Hammerly introduces the topic of AI dev tools and guest speaker Tomek Porozynski.
- 00:16:08 · Speaker Introduction — Tomek Porozynski
- Tomek introduces himself as a Staff ML Engineer and Google Developer Expert from Poland.
- 00:28:18 · Project Showcase: Multi-Voice Audiobook — Tomek Porozynski
- Tomek describes an exciting project he built that uses Gemini Text-to-Speech to convert text into a multi-voice audiobook.
- 01:47:29 · Tools and Technology — Tomek Porozynski
- He explains using Gemini CLI with skills for Vertex AI and the Gemini API, and how he moved the project to a Google Colab notebook.
- 02:33:29 · Architecture Deep Dive — Tomek Porozynski
- Tomek details the application’s architecture, which involves a combination of serial and parallel processing to generate the audio.
- 04:26:24 · Challenges and Learnings — Tomek Porozynski
- He discusses the challenge of rapidly evolving AI tech and how using ‘skills’ helps models access up-to-date information and APIs.
- 05:48:29 · Getting Started with AI Dev Tools — Tomek Porozynski
- Tomek advises starting small and recommends Antigravity IDE for coders and Google AI Studio for those without coding skills.
- 10:14:04 · Conclusion — Aja Hammerly
- Aja thanks Tomek for sharing his insights and details about his open-source project.
Demos (1)
- 00:38:09 ✓ · Multi-Voice Audiobook Generator — Tomek Porozynski
- Tomek verbally described a workflow he created that uses Gemini Text-to-Speech to analyze a story, identify characters, assign different voices, and generate a multi-voice audiobook. He also mentioned it’s an open-source project available as a Google Colab notebook.
Notable Quotes (6)
- 00:38:09 — Tomek Porozynski:
I’m really excited about Google, or Gemini Text-to-Speech APIs.
- 02:46:11 — Tomek Porozynski:
I open-sourced the solution and I used that on the DevFest talks to actually show people how that can be done.
- 05:33:28 — Aja Hammerly:
I don’t know what your experience has been like, doesn’t always go exactly the way I planned. I always run into something.
- 06:38:00 — Tomek Porozynski:
My advice would be not to give up when the first version is not perfect.
- 07:00:23 — Tomek Porozynski:
One thing that I think is sometimes overlooked is the plan phase or the brainstorming phase.
- 09:07:05 — Tomek Porozynski:
I think the great idea is to check Google AI Studio.
Visual Signals
On-screen (4)
- 00:00:00 ·
Google Cloud Next 26 logo- Brands the event and the year.
- 00:33:21 ·
Aja Hammerly, Director DevX AI, Google Cloud- Identifies the speaker and her role.
- 00:21:20 ·
Tomek Porozynski, Google Developer Expert, Cloud & AI- Identifies the speaker and his role.
- 10:29:26 ·
Google Cloud Next 26 logo- End card for the video segment.
Stage (1)
- 00:00:00 · The video begins with two speakers, Aja Hammerly and Tomek Porozynski, seated at a desk in a studio/booth setting at the Google Cloud Next conference.
Key Topics
AI Dev Tools · Gemini API · Text-to-Speech · Google AI Studio · Gemini CLI · AI Agents · Agent Skills · Generative AI Development · Proof of Concept (POC) · Google Colab · Open Source AI · Developer Experience (DevX) · Prompt Engineering · Application Architecture · Rapid Prototyping
Takeaways
- Gemini Text-to-Speech API is powerful enough to create sophisticated applications like multi-voice audiobooks by identifying characters and assigning unique voices.
- For rapid prototyping, Gemini CLI combined with ‘skills’ is an effective tool for testing ideas and building proof-of-concepts.
- The development process for AI applications should include a ‘planning phase’ where developers converse with an AI agent to brainstorm, refine ideas, and select the tech stack before writing code.
- Google offers accessible entry points for all skill levels: Antigravity IDE for developers who prefer coding, and Google AI Studio for a no-code/low-code approach to building and deploying AI apps.
- The fast pace of AI innovation is a major challenge; using ‘skills’ with AI agents is a key strategy to ensure models can access the most up-to-date information and APIs.
- Open-sourcing projects and using shareable formats like Google Colab notebooks are excellent ways to contribute to the community and help others learn.
- It’s possible to go from an idea in Google AI Studio to a fully deployed application on Cloud Run with a custom domain in a very short amount of time.
- Iterative development is key when working with generative AI; the first result is rarely perfect, and developers should be prepared to refine and troubleshoot.