

Is Google's new Image Model OVERHYPED ? (Test Results)
4 snips Aug 29, 2025
Is Google's latest image AI a game changer or just hype? The hosts test Gemini 2.5 Flash, revealing a significant jump in performance but notable flaws, like context loss after a few generations. They explore its effectiveness in YouTube thumbnails and Facebook ads, alongside critical safety concerns. Plus, they introduce two useful new tools: PicShop and Co-Drawing. The discussion delves into AI's impact on advertising and how it compares to traditional methods, raising the question of whether it's worth switching from current AI workflows.
AI Snips
Chapters
Transcript
Episode notes
Progress Without Perfection
- Gemini 2.5 Flash is a meaningful step forward but not a complete revolution.
- It still makes basic context and common-sense errors after multiple edits.
Factor Cost Into Model Choice
- Use the Gemini API for cheaper image generation at about $0.04 per image.
- Consider cost differences vs OpenAI when choosing a production workflow.
Thumbnail Mashup Worked—But Was Editing, Not Recreation
- Mark combined an Ali Abdaal thumbnail screenshot with his studio photo and got a convincing edit.
- Gemini often edits (cuts) the uploaded subject rather than fully regenerating them, preserving likeness.