Clarifai Visual Search vs Moderation API

Clarifai Visual Search lets people locate people, places, objects, topics, and products in images and videos without describing them with words. In ecommerce, shoppers can submit a photo to find similar products, identify purchasable items within rooms or outfits, and receive alternatives when an item is unavailable. Search can rank whole images by visual similarity or use detected regions for localized searches. The API supports semantic similarity retrieval, AI-generated vector embeddings, metadata filters, and combined conditions. One object-search workflow accepts SVG, PNG, JPG, or GIF files up to 800 by 400 pixels. Deployment is available through API, web, or self-hosted arrangements, including shared or dedicated cloud, a customer VPC, on-premise, and edge environments. Clarifai lists object detection, image classification, face recognition, OCR, and custom models. It also lists integrations such as LangChain, LlamaIndex, Databricks, Unstructured.io, CrewAI, and Vercel. Clarifai says customer data is private by default, is not used for training unless shared with the Community, and is covered by SOC 2 Type 2 compliance.

Moderation API is a content moderation service that combines an API with a review platform for text, images, video and audio. It supports more than 120 languages and offers over 20 pre-built detection models for issues such as toxicity, personal information, spam, self-harm, scams and hate speech. Teams can describe their own moderation rules in plain language; models apply them with explanations and confidence scores. Review queues and author moderation support human decisions, and the product says team reviews help train its AI. It also considers conversation history when assessing a message. The service lists integrations with tools including Discord, Airtable, Shopify, WordPress and Zapier, plus server-side SDKs for TypeScript/Node.js, Python, PHP and Ruby. A free Hobby plan includes 100 messages per month; Basic is $79 a month. Usage units depend on media: text costs one unit, audio two, images three, while video starts at three and varies with length and enabled models. The maker says content is not stored or accessed by people.