Bob vs Gemini

Gemini's surface is breadth. It takes text, images, audio, and video as input, speaks back in voice conversation, and pulls live results from Google Search while grounding answers in Gmail, Docs, Calendar, Drive, Maps, and YouTube. Current model tiers are split by workload — 3.7 Flash for reasoning and multimodal tasks, 3.6 Flash for speed, 3.5 Flash-Lite for lightweight jobs, 3.5 Flash Cyber for security-focused work — plus Gemini Omni 1.1 Flash for video (10-second scene analysis, 360p draft, 40-second extension) and 3.5 Transcribe for audio. Computer-use actions let it click, type, and navigate a GUI, and the Gemini API exposes all of it to developers.

Bob's surface is depth on one narrow job: getting text off your Mac screen. It offers three capture modes — select-text translation, screenshot-region translation, and manual input — and three OCR entry points: screenshot OCR, silent screenshot OCR (no popup, back-to-back captures), and Finder image-file OCR. 1.20.0 added standalone OCR (OCR without going through translation first) and original-image translation via an OCR plugin. Offline OCR works without a network; continuous recognition merges multi-segment captures with smart paragraphing; QR codes are read from screenshots. On the translation side it aggregates 30+ engines — DeepL, Google, Microsoft, Amazon, Baidu, Youdao, Tencent, Volcengine, Apple — and can run several at once side by side, with a bundled concise English-Chinese dictionary.

Verdict on capability: no overlap worth comparing. Gemini creates and reasons across media; Bob extracts and converts text in-app.

Both list as freemium, but the economics are different animals. Gemini is a bundle: you get the assistant woven into Search, Gmail, Docs, Calendar, Drive, Maps, YouTube, Chrome, and Google Home, with usage included at whatever tier you're on. Your real cost is the Google subscription or API consumption — there's no per-feature meter for the Workspace integrations, and the API is the only line item developers see directly.

Bob is free to download from the Mac App Store, but that's the front door, not the bill. Most of its translation and OCR services require you to apply for and supply your own API keys, and those providers bill you per character, per request, or per usage tier. So your actual spend depends on which engines you enable and how much you translate — heavy DeepL or cloud-OCR use is a real recurring cost, while offline OCR and the bundled dictionary cost nothing extra. The trade-off is control: Bob's advanced users choose their engines and own the keys, which is why the "not for" list includes people who want zero configuration.

Practical read: Gemini's cost is decided by your Google plan; Bob's is decided by your translation volume and engine choices. There is no meaningful per-seat comparison between a bundled assistant and a bring-your-own-key desktop utility.