Prepare your data

Do this part first, before anything else. Requesting an export is often the slowest step in the whole process and it is entirely out of your hands once started.

Longest wait for an export

30days

Spotify's extended streaming history can take this long to arrive.

from Start requesting exports today

A Google Takeout with Photos

40–50GB

Common size once expanded; the appliance needs room for the expanded copy.

from A note on space

Largest single file the web upload takes

10GB

The appliance gateway rejects a larger request. Copy bigger files into the inbox folder instead.

from A note on space

Start requesting exports today

Some services hand you a file immediately. Others take days, and a few take weeks — Spotify’s extended streaming history can take up to 30 days.

Every source in the catalogue on the Import page shows how long its export usually takes — as a badge from minutes to weeks — alongside the steps to request it. Work through the ones you care about and request them all now; they will arrive while you do something else.

More sources make a better memory

Adding more data imports compounds in value, because different sources describe the same life from different angles: a photo says where you were, a calendar says who you were with, a message thread says what it was about. Any one of those alone is thin.

A good starting mixture:

• Contacts — a vCard .vcf or a contact-list CSV. Import this early: it is what gives everyone else a name instead of an email address.
• Messages — iMessage, an Android SMS backup, Gmail, Telegram, WhatsApp.
• Location history — Google Takeout location history, GPX tracks, GeoJSON.
• Photos — Apple Photos, Google Photos, or any folder of media. Photos carry dates and often coordinates, so they anchor a lot of the rest.
• Calendar — an .ics export.

Keep your originals

Keep the archives you downloaded, even after a successful import.

This importer is under active development. An upgrade may improve how a source is read, and getting the benefit means importing that archive again. Some services will not let you request a second export of the same period, and a few will not give you back data you have since deleted from them.

When you upload an export by drag and drop, or with Choose files or Choose folder, your browser sends a copy and your original stays where it is. The appliance deletes its copy as the import consumes it. If you copy files into the inbox folder, the appliance treats them as copies too, and deletes them as their import consumes them. Copy your exports there instead of cutting and pasting, so that your originals are not deleted.

Gather the exports in a well-named folder

Most exports arrive in your Downloads folder as .zip files, sometimes several per request when a service splits a large archive. Some may come as .csv or .xml files. Placing them all in a folder named something like “CI Memory Data Exports” can help differentiate them from other downloads, especially since many don’t have particularly helpful names.

You do not need to unzip anything. Upload the .zip, .csv, or other file exactly as it downloaded — Memory expands it, including archives nested inside other archives, which is how some sources such as Google Takeout ship your data.

A note on space

Memory keeps its own copy of photos, videos and documents for search, so an import can need a lot of disk. Google Takeouts that include your Photos backups can run 40–50 GB or more once expanded. Four checks and limits apply.

Working space for the expanded archive. Before the Unpack stage expands anything, it adds up the uncompressed size of your archives, subtracts the archives themselves (it deletes each zip as soon as it has expanded), adds 1 GiB of headroom, and compares the total with the free space on the volume that holds the temporary working folder. If the space is not there, the import stops with a “not enough disk space to unpack” message that names both figures, and nothing is written. Free space and start the import again.

Database space. Unpack also refuses to start when the volume that holds the database has less than 2 GiB free, because the records could not be saved afterwards.

Space for stored files. Before the Store files stage copies photos, videos, voicemails and attachments into Memory’s file storage, it adds up the declared size of that media. It needs half that total plus a reserve of 2 percent of the volume (capped at 100 GiB) free on the same volume as the working folder. If the space is not there, the stage fails with a “not enough disk space to store this import’s media” message, uploads nothing, and waits for you to free space and select Try again. The check measures the working folder’s volume as a stand-in for the file storage, and it counts only sizes the archive declares, so leave more free space than it asks for if your archive is mostly media.

Size limits. The app accepts a single uploaded file of up to 50 GiB, and expands an archive up to 100 GiB and 2,000,000 entries. The gateway in front of the app limits a single browser upload to 10 GB, though, so a file over 10 GB fails to upload from the wizard. Copy it into the inbox folder instead, or request the export again with a smaller part size. The Gmail and Google Photos guides on the Import page suggest 50 GB parts; a part that large needs the inbox. An operator can change the expansion limits with the CI_IMPORT_MAX_ARCHIVE_BYTES and CI_IMPORT_MAX_ARCHIVE_ENTRIES environment variables.