Product guide · Version 0.3.0.32
What Token Saver does: lighter files for your AI assistant
By Genoskwa Enterprises Inc. · Updated (America/Toronto)
Token Saver is a desktop app for Windows and Mac that prepares lighter versions of your own files before your AI assistant reads them. It transcribes audio and video locally, makes smaller video or image copies, and extracts text from PDFs and other supported documents. You choose what to send or share; the originals stay on your computer.
This guide covers the features in public version 0.3.0.32, available for Windows 10/11 (64-bit) and Mac with Apple silicon. Token Saver is an independent product from Genoskwa Enterprises Inc.
Turn a local recording into a transcript
Drop a video or audio file into the app to get a text transcript made on your computer. Your assistant can work from the words without receiving the original recording.
When you press Send, Token Saver attaches the transcript. If you also selected “Also make a smaller video” and that smaller copy exists, it is attached too. The original recording is not attached. Audio files such as WAV stay transcript-only.
Make a smaller copy when you need the visuals
For a big video, you can choose to make a lighter copy locally. Token Saver can also make smaller images. A smaller file is not a measured token-saving percentage: file size and what an assistant needs to read are different things. Choose the version that contains the information needed for your task.
Extract document text instead of sending everything
Token Saver pulls text from PDFs, docs, and other supported files. Your assistant can work from an outline, selected PDF pages, or a slim text extract rather than a full document. The feature overview describes the current capabilities.
Read changed spreadsheet rows
Change tracking lets you ask for the spreadsheet rows that changed instead of re-reading the whole sheet. It is a way to focus the next request on the update, rather than sending the same unchanged material again.
Share only folders and extracts you approve
Choose folders in the app and approve each sync, or select Always for a particular folder. Secrets such as .env files are skipped. The hosted MCP connector uses OAuth and the same email as the app, with no API key to paste. It works with Grok Bot through that hosted connector.
Raw recordings, full documents, inboxes, and full memory stay on your device. A transcript, chosen smaller copy, slim extract, or change list can leave after you approve sharing or press Send. Approved hosted extracts expire after at most seven days. Sign-in, device registration, licence checks, and sync still involve connection data; “local” does not mean the app never communicates with a service. See the privacy details.
Reuse shared memory and cached answers if you choose
You can optionally mark individual memory facts and previously approved answers as shared. Only those selected facts or answers leave; full memory stays on your computer. This gives the assistant approved information to reuse without reconstructing it each time.
Understand the savings counter
The counter gives a rough, character-based estimate of tokens avoided. Its estimated dollar figure uses the public API input rate of $2 per million tokens. It is not an actual billing measurement, a measured percentage, a guaranteed saving, or a SuperGrok refund. Read the savings-counter FAQ before interpreting the number as money.
Start with one file you already have
- Download Token Saver 0.3.0.32 for your supported computer and read the install notes.
- Drop in a recording or supported document, or select a folder you approve.
- Choose the transcript, smaller copy, or extract you want your assistant to use, then approve sharing or Send.
Early Access checkout starts inside the app after sign-in, under Upgrade. There is no separate web checkout. The homepage pricing section has the current terms.