recstrip
New in v0.20 · the research toolkit

Every interview, turned into evidence — on your machine.

recstrip transcribes your recordings, knows who said what, and turns 20 sessions into themes, highlights and reels — with every claim linked back to the exact moment it was said. Local by default. No account.

Download for Windows Watch the 21s film

Free · Windows 10 / 11 (x64) · latest release

recstrip — Acme discovery · 22 sessions
Local by defaulttranscription, speakers & search on your CPU
Knows who spokespeaker labels you can rename
Cited answersevery claim plays its moment
No accountnothing to sign up for

The film

Twenty-two interviews. One question. Twenty-one seconds.

22 interviews.

One question.

Evidence you can point at.

v0.20

From recordings to a research repository

The workflow research teams pay enterprise plans for — speakers, coding, reels, sharing — built into a free app that keeps your interviews on your own computer.

01

Knows who said what

Speaker detection runs on your computer, or in one cloud pass together with the transcript. Rename a speaker once, merge two that are the same person, link them to a participant — and see who did the talking.

  • Local (sherpa-onnx) or cloud (Whisper + pyannote)
  • Talk time, turns and longest monologue per speaker
  • Names flow into insights, highlights and exports
02

Highlight, tag, and sort

Select a quote and press H. Tag it from your project's codebook, or let the AI propose highlights for you to keep or dismiss. Then sort everything on an affinity board — drag a card to re-tag it.

  • Codebook with groups, colours and definitions
  • AI-suggested highlights, never applied silently
  • Themes show “raised in 7 of 22 sessions” — and who disagreed
03

Clips and reels for stakeholders

Every highlight on audio or video is a clip. Line them up into a reel and render one MP4 — each clip labelled with its session, the transcript burned in, and names or account numbers bleeped when you need to.

  • Trim clips a second at a time
  • Audio-only interviews get a clean caption frame
  • Plays straight through inside the app
04

Ask from Claude, Cursor or any AI assistant

recstrip includes a local, read-only MCP server. One click in Settings connects Claude Desktop or Cursor; your assistant can then search your sessions, read transcripts, themes and highlights, and cite them — without an upload or an account.

  • Semantic search that understands meaning, across languages
  • Opens your data read-only; nothing leaves the machine
  • Works with Claude Code: claude mcp add recstrip

Everything a meeting leaves behind, made searchable

Built for people who need the record to be accurate — closer to a research tool than a chatbot.

Runs entirely on your machine

Whisper transcription and the AI models run locally and are cached on your device. No account, no upload, no cloud dependency for the core workflow.

Answers you can verify

Ask anything across a project's sessions. Each answer shows its sources first and links every citation to the exact timestamp in the transcript.

Finds the decisions

Automatically extracts decisions, requirements, action items, risks, and open questions, then groups them into themes across sessions, so you see that a point came up in 7 of 22 interviews, and who disagreed.

Reads the shared screen

Keeps the keyframes that matter — slides, docs, dashboards — and skips the webcam feeds. On-screen text is OCR'd so it's fully searchable.

More than recordings

Import video, audio, PDFs, and emails into one project. It all becomes part of the same searchable, ask-able knowledge base.

Knows who said what

Speaker detection on your computer, or in one cloud pass with the transcript. Name speakers, link them to participants, and see each person's talk time.

Highlight, tag, clip

Select a quote to highlight it, tag it from your codebook, sort it on an affinity board, and turn highlights into a video reel for stakeholders.

Works with your AI assistant

A built-in, read-only MCP server lets Claude, Cursor and other assistants search and cite your sessions — locally, without an account.

Share without uploading

Export a stakeholder report, Word transcripts, subtitles, CSVs or a Markdown folder — with names, emails and account numbers redacted when you need it.

Enterprise features. Without the enterprise plan.

Research repositories and meeting notetakers often reserve these for Business or Enterprise tiers — and run them on their servers. recstrip runs them on yours.

Typical cloud toolrecstrip
Speaker detectionincluded, in the cloudlocal or one cloud pass
Themes across sessionsoften a paid tierincluded
Custom analysis templatesoften a paid tier5 built in + your own
PII redactionoften enterprise onlyexports + before cloud AI
Highlight reelsvaries by planrender MP4 locally
AI assistant access (MCP)via their cloudlocal, read-only
Where your recordings livetheir serversyour computer

How it works

  1. 1

    Import your sessions

    Drop in meeting videos, audio, PDFs, emails or Word files. One project per study or engagement.

  2. 2

    It does the grunt work

    Transcribe, label speakers, read the shared screen, extract insights, suggest highlights and build a semantic index — locally.

  3. 3

    Ask, code and share

    Chat with every session, tag highlights, spot themes, then export a report, a reel or hand it to your AI assistant.

Private by design

Your recordings, transcripts, highlights and API keys stay in a local folder on your computer. recstrip only touches the network to download its models, and — only if you choose — to transcribe on Replicate or to call your own LLM provider. Turn on pseudonymisation and names, emails and account numbers are replaced before any text leaves your machine, then restored in the answer. There is no recstrip account and no telemetry.

Get recstrip

Free, and yours to run offline.

Download for Windows

Windows 10 / 11, 64-bit · latest release · first launch downloads the local models (~2 GB, one time)