Import Media

Import Media turns the videos, photos and recordings from one day or event into a single entry. With a Gemini profile, the video’s picture and sound are read together, and frames from the moments that matter are attached to the entry.

Starting an Import

• iOS: In the journal list, open the More menu and choose Import Media. Add items with Add from Photos or Add Files.
• macOS: Choose Import Media from the More menu and use Add Files, or drop several files on the main window. Anything that is not a video, photo or audio file is ignored.

You can add as many items as you like. Each row shows a frame, the time it was recorded and its length, and notes No sound or Time unknown where that applies. Swipe a row, or use Remove from its context menu, to take it out.

Order, Date and Place

Items are listed in recorded order: dated items by the time they were taken, then undated ones in the order you added them. This is also the order the model reads them in.

The entry is dated and placed where the first item was recorded. The date comes from the video’s metadata, the photo’s EXIF data, or the audio file’s creation date. The location comes from the first item that has GPS data. The import never uses your current time or place.

What the Entry Contains

The entry’s attachments follow the same order as the items:

Item	Attached
Video	Selected stills (JPEG) and the video’s audio (M4A). The original video too, if Attach Original Videos is on
Photo	The original photo
Recording	The audio file

Attach Original Videos is off by default, which keeps your journal folder small. The originals stay where you picked them from. The toggle shows how much space the videos would take.

Choosing a Profile

What happens depends on the profile you pick.

Gemini profiles

A profile that uses a Gemini model with a prompt reads everything: the picture and sound of each video, and each photo. It writes the entry from what is said and what is seen, following the profile’s prompt, and picks up to six moments whose frames are attached.

To keep uploads small, Ayumi sends a low-resolution copy of each video (360p, one frame per second) and photos scaled down to 1536 px. Originals and full-size stills stay on your device. Large imports are uploaded through the Gemini Files API, and the uploaded copies are deleted afterwards.

Sending videos and photos needs its own one-time consent for each provider, separate from the consent to send audio, because they can show faces, other people and places.

Other engines

On-device transcription and the other cloud engines (Gemini Transcribe, Muse, Grok, Groq) transcribe sound only. They can be used when exactly one item has sound. About four stills, spread across the videos, are attached. On-device, nothing leaves your device; with a cloud engine, only the recording’s sound is sent.

Note

Photos alone, or videos without sound, need a Gemini profile, since only Gemini can read pictures. Combining several recordings also needs a Gemini profile.

After You Import

The import is added to the transcription queue like any recording, and the sheet closes right away. If it fails, retry it from the queue. If you create the entry without a transcript, a video import gets a placeholder title with the date.