Get the gist.
Keep the good parts.
Turn available YouTube context and readable web content into useful summaries, without losing the thread.
Your AI companion for YouTube and the web. Turn long videos, pages, and documents into clear answers—without leaving what you’re exploring.
A long watch. A dense read. A question that won’t leave you alone. Start wherever curiosity takes you.
Turn available YouTube context and readable web content into useful summaries, without losing the thread.
Ask follow-up questions about the page or video instead of starting your research from scratch.
Click timestamped YouTube answers to jump to the moment, when source timestamps are available.
Explore supported PDFs and spreadsheets, or ask about a visible-tab image with a compatible vision model.
Optional read-aloud and local speech features; some need separate models or a local helper.
Blip meets you in the flow of browsing. Choose your AI connection, open something interesting, and start a conversation. For an authorized remote model server, connect through a secure tunnel to a localhost port you configure.
In Settings, choose API for Gemini, OpenAI, OpenRouter or NVIDIA with your key. Choose Local for Ollama or LM Studio, check the connection, select an installed model and Save. Ordinary local servers need no API key.
Launch the Blip bubble on YouTube, or activate it from the toolbar on another supported page.
Get a summary, ask questions, explore available timestamps, save conversations, or export chats.
Ask questions about a Google Sheet you’re working with. Connect your account inside Blip to bring selected spreadsheet content into the conversation.
Read the data policyBlip has no developer-operated AI backend and does not collect usage analytics. Your extension settings and chat history stay in your browser. API requests go directly to your chosen provider. Local mode connects directly to your Ollama or LM Studio localhost endpoint, without a Blip server.
API providers receive relevant content under their own policies. Local mode uses downloaded models on your device, or an authorized server reached through your own tunnel. Tunneled prompts leave your device for that server; localhost alone cannot verify where inference runs. Known cloud-backed models are blocked, with no automatic API fallback. Page retrieval and model downloads can still use the network.
From the first question to the next big idea, Blip keeps the conversation going.
Add to your browser