Summary Google ML Kit is a free mobile SDK for on-device AI on Android and iOS, with vision, language, and Gemini Nano-powered GenAI APIs. Review features, limits, and best use cases. Google Vision AI covers Cloud Vision API, Document AI, Vertex AI Vision, and related image understanding tools. Review features, pricing, and which Google vision product fits. Amazon Rekognition is an AI image recognition platform for developers, ML teams, ecommerce teams, security teams, and media platforms. Compare features Pricing Model Free Freemium Paid Starting Price 100% Free Free + from $1.50/per 1,000 Vision API units Paid Current Version — — — Category Rankings Strengths Free entry point for production mobile AI features Strong fit for offline and privacy-sensitive app workflows Broad API coverage across vision, language, and on-device GenAI Easier to integrate than building custom mobile ML pipelines from scratch Broadest product coverage in the Google Cloud computer vision stack Strong fit for teams that may grow from OCR into document or video workflows Cloud Vision API still offers a low-friction starting point Good alignment with broader Google Cloud data and AI infrastructure Focused AI image recognition platform workflow gives users more structure than a generic assistant. Public product messaging makes the main use case clear enough for fast evaluation. Useful for repeat work where templates, exports, integrations, or review steps save time. Can reduce blank-page or manual production work while keeping humans in control of final quality. Limitations Best suited to mobile apps, not general web or backend AI products Capability depends on supported devices, Android versions, and packaging choices Less flexible than building fully custom models and serving logic yourself Gemini Nano-backed features are narrower than cloud-scale model platforms Not a single product with one clean pricing model Buyers have to choose among multiple overlapping Google products Some use cases are better served by sub-products than by this top-level page Costs become harder to estimate once workloads span several services Public pages may not expose every limit, integration detail, security term, or procurement requirement. Output quality depends heavily on input quality, source material, prompts, and review discipline. Highly custom workflows may still require specialist review, manual editing, or additional tools. Advanced exports, admin controls, API access, or commercial rights may require paid or enterprise plans.