Skipped-passage detection
Detects fast scroll gestures and estimates the skipped range instead of summarizing the whole page, so the recap stays tied to the passage you moved past.
Skim Recap
Skim Recap is a Chrome extension that returns a short local recap of the passage you scrolled past, helping readers keep their place in long articles, docs, and research writing. It uses Gemma 4 E4B through LiteRT-LM on-device and keeps the recap beside the cursor.
What Skim Recap does
Skim Recap is a Chrome extension that catches passages you scrolled past and returns a short recap beside your cursor. It is built for readers who want to skim long articles, documentation, essays, and research posts without losing the thread of what they skipped.
The product uses Gemma 4 E4B through LiteRT-LM in a Chrome offscreen document with WebGPU, and the source pages say the skipped text is summarized on-device rather than sent to a hosted LLM API. A fast scroll triggers the capture, the skipped range is extracted, and the recap panel appears near the cursor so the reader can stay in place.
Skim Recap exposes a small popup for enabling the extension, setting the flick threshold, and preloading the model. It also offers two recap views from one generation, plus on-device translation and a later release path for selecting a term and asking for a Feynman-style explanation when a recap alone is not enough.
Core features
Cursor-native recap panel
Displays the recap beside the cursor in a compact panel, keeping attention in place while the text is returned.
Popup controls
Runs from a popup with an enabled toggle, adjustable flick threshold, and a model preload action for starting and tuning the workflow.
Dual recap layouts
Uses one local generation for two display modes: Focus shows one paragraph, while Smart formats the same output as numbered points without re-running the model.
On-device language handling
Supports on-device translation and can also generate the recap directly in a selected language in a single pass.
Reusable local model
Caches the model locally after download so the same browser session can reuse it across tabs, with the model remaining available until Chrome restarts.
Where Skim Recap fits
Skimming long articles
Use it when you move through a long-form article quickly and want the omitted paragraph returned without leaving the page or opening a separate summary view.
Following dense documentation
Use it while reading technical docs or essays where missing one passage can break the thread of the argument or the procedure.
Recovering skipped research context
Use it for research posts and analytical writing when you want a brief recap of the exact section you passed over, not a summary of the whole page.
Reading in another language
Use the translation action or direct language generation when you need the recap in another language while staying in the same reading flow.
Explaining unfamiliar terms
Use the Feynman selection flow introduced in v0.4 when a recap names an unfamiliar term and you want an explanation on a specific word or phrase.
Pros and Cons
Pros
- Keeps the recap anchored to the specific passage you skipped instead of summarizing the entire page.
- Runs locally with Gemma 4 E4B through LiteRT-LM and WebGPU, reducing dependence on a hosted LLM API.
- Places the result beside the cursor, which helps preserve reading position.
- Lets you switch between Focus and Smart without regenerating the answer.
- Includes on-device translation and the option to generate directly in a chosen language.
- Caches the model locally so the same browser session can reuse it across tabs.
Cons
- The recaps are intentionally bounded to the skipped passage, so they will not explain concepts that the page never defined.
- The first model download is large at 2.97 GB, according to the product page.
- It is a Chrome extension, so the documented workflow is centered on Chrome rather than broader browser support.
FAQ
How does Skim Recap get started?
Skim Recap is activated from a small popup with an enabled toggle, an adjustable flick threshold, and a model preload action. Once active, it watches for fast scroll gestures and shows a recap beside the cursor after scrolling settles.
Does Skim Recap run locally or in the cloud?
It is built as a Chrome extension and runs Gemma 4 E4B locally through LiteRT-LM in a Chrome offscreen document using WebGPU. The product pages say the skipped text is summarized on-device and not sent to a hosted LLM API.
What formats does the recap output support?
The generated recap appears in two layouts: Focus, which presents the recap as one paragraph, and Smart, which formats the same generated text as numbered points. Switching between them does not regenerate the model output.
Can I translate a recap?
Yes. The panel includes a translation action, and the product page says translation runs on-device as well. The recap can also be generated directly in a chosen language in one pass.
What is Skim Recap best suited for, and what is one limitation?
The product is designed for long articles, documentation, essays, and research posts. The about page also notes a limitation of bounded recaps: if the page used a term without defining it, the recap cannot explain that outside context until the v0.4 Feynman selection flow is used.
Quick Facts
- Category
- Chrome extension / AI reading tool
- Primary workflow
- Fast scroll detection followed by a local recap of the skipped passage
- Model
- Gemma 4 E4B Instruct through LiteRT-LM
- Runtime
- Chrome offscreen document with WebGPU and local WASM runtime assets
- Platform
- Chrome
- Source domain
- skim-recap.vercel.app
Alternative a Skim Recap
Notewarp
Notewarp is an AI note-taking tool that turns voice notes, uploads, or pasted text into cleaner notes, transcripts, summaries, and action items. It helps people capture rough ideas quickly and then edit, export, or share the result.
QuickQuill
QuickQuill is a macOS dictation and transcription app that runs locally on the device. It helps users record meetings, transcribe audio, generate summaries, and export notes without using a cloud service.
MeetStream
MeetStream is a meeting bot API for Zoom, Google Meet, Microsoft Teams, and Webex. It helps developers record, stream, and analyze meetings programmatically, with usage-based pricing and a $5 free credit to start.
AI PDF Summarizer
Uno strumento alimentato da IA che riassume rapidamente ed efficientemente documenti PDF, fornendo informazioni chiare e consentendo agli utenti di interagire con il contenuto.
Tactiq
Tactiq is an AI note taker for Google Meet, Zoom, and Microsoft Teams that transcribes meetings live and turns them into summaries, action items, and follow-up outputs. It is built around a Chrome extension and supports team workflows through sharing and integrations.
心流AI助手
心流AI助手是一款面向知识获取与内容处理的 AI 助手，可用于实时问答、长文总结和写作辅助。公开页面显示它覆盖日常百科、生活信息和学术论文相关场景，并提供 App 体验入口。