veridive

Veridive searches spoken content on YouTube, podcasts, lectures and interviews and returns answers cited to the exact second the claim was spoken. It supports live deep search, chat over a defined set of sources, image-question input and multilingual queries, with a creator tools area and an MCP endpoint for external clients.

Veridive is an answer engine built for spoken content rather than written pages. Instead of returning a list of links to articles, it searches video and audio sources such as YouTube clips, podcasts, lectures and interviews, then returns a synthesized answer with each claim traced back to the precise moment in the recording where it was said. The pitch on the homepage is direct: think ChatGPT, but every answer linked to videos. The workflow has two entry points. DeepQuery runs a live search across YouTube and other spoken sources, so a question in plain language comes back with citations that can be played from the correct timestamp. DeepContext Chat works against a smaller, defined body of material, letting a user query a specific set of videos and images instead of the open web. There is also an image input, so visual questions can seed a search. The product describes itself as multilingual, and results are meant to work across languages rather than only English sources. The homepage illustrates the use case with curated threads assembled from dozens or hundreds of scanned videos. Examples include a step by step phone screen replacement built from 53 scanned videos, a Paris tourist trap guide drawn from 91, a mattress buying guide from reviewers who are not sponsored, and a collection of highlights from Lex Fridman's guests built from 47 videos. These examples point to the intended audience: people who want practical or research answers and care about whether the source is a paid promoter, plus anyone who prefers hearing a claim in context over reading a summary. Beyond search, Veridive exposes a creator side. The Agents area (Deep Watch) and Tools sections suggest workflows for building curated collections, and an MCP page indicates a protocol endpoint for connecting the engine to other AI clients. A library, chats and creatives area round out the account dashboard. Pricing tiers exist on a separate page. Against general chatbots, the differentiator is verifiability: the answer is only as good as the transcript, but the user can check it in one click.