Hacker News new | ask | show | jobs
by aaronkwhite 164 days ago
Started messing with Gemini 3 Pro image generation over the holidays. First tried having Claude interact with the Gemini CLI — worked fine, but burned through tokens fast.

Built a UI instead. Hits the Batch API directly, which is half the cost of real-time: text-to-image, image-to-image, and queue management. NextJS, SQLite, everything local, bring your own API key.

~$0.02/image at 1K, $0.07 at 2K, $0.12 at 4K.

Built it with Claude. Codebase is simple enough that any agent can help iterate on it. Planning to add Veo 3 when Google opens batch access for video. Feedback welcome.