Recordings: Voice and Video Notes

A recording is an audio or video note you made, together with the text of what was said in it. Nocarta turns your speech into text in the background, so you can dictate an answer while filling a form, or narrate a video and later turn what you said — and what the camera saw — into structured records.

Find your recordings in the sidebar under Recordings.

Three Ways to Create a Recording

Dictate while filling a form (audio). In chat-mode filling, a microphone button appears next to the send button whenever a question is waiting for your answer. Tap it, speak, tap again to stop — the spoken text is offered as the answer.
Record a video while filling a form. The same row has a Record video button. Video recordings are processed in the background and appear under Recordings.
Upload a file you already have. On the Recordings page, use the card "Add a recording you already have": pick the file, optionally type the language, and click Upload. You'll see: "Recording uploaded. Transcription is running."

The microphone and camera buttons only appear when speech-to-text is available on your account, and only while a question is waiting for an answer.

Supported Files and Size

Audio: MP3, WAV, OGG, WebM, MP4/M4A, AAC
Video: WebM, MP4, QuickTime (.mov — the usual phone format)
Size: depends on your plan — the upload card shows your exact limit ("Audio or video, up to N MB"). Over the limit you'll see: "Recordings must be smaller than N MB."
Whether a file counts as audio or video is detected from the file itself.

What You Get After Processing

Processing runs in the background. A recording moves through these states: Waiting to be processed → Processing… → Transcribed (or Could not be processed).

On the recording's page you'll find:

A player to listen or watch
Status, Length, and Language
Attached to — the form it belongs to, or "Not attached to anything" for a standalone recording
The full transcript, with line breaks preserved
Word timings — when available, each word's exact position in the recording ("N word timings available"). Automations can use these to find the moments where you said something.
Frames (video only) — still pictures sampled from the recording, each labelled with its position in seconds. Frames are taken roughly every two seconds, spread across the whole video. If one moment can't be turned into a picture, that frame is simply skipped — the transcript is not affected.

Choosing the Language

The spoken language is detected automatically. For short clips, detection can guess wrong — typing the language code (for example en, it, pt) in the Language box next to the upload button measurably improves accuracy. The transcript is always in the language you spoke; it is not translated.

The Recordings Page

Overview tiles: how many recordings you have, storage used, total length, and how much space sits in the trash
Filters: All / Audio / Video
Each row shows the transcript (or its status while processing), the date, the duration, what it's attached to, and a delete button

Deleting and Restoring

Deleting moves a recording to the Trash section — "Delete this recording? You can restore it during the retention window."
Deleted recordings are kept for 30 days, then permanently removed. Until then, a Restore button brings them back.
Storage space is freed only after the retention window — both the Recordings page and the Usage page show trash space separately.
Recordings are never shared: even when attached to a shared form, only you can open your recording.

Using Recordings with Automations

Once a recording has a transcript, its page shows "Run an automation on this recording". Your automation decides what to do with what was said — where to split it, and what to create.

Automations that run in the browser are labelled "Runs here — no desktop needed" — so a video recorded on your phone can be processed on that same phone.
Automations that need the desktop app are shown too, labelled "Needs the desktop app".

What an automation can do with a recording:

Read the transcript and its word timings, and split one narrated recording into several records — for example, one record per time you said a marker word of your choice. The marker word is yours: nothing in Nocarta decides it for you.
Attach a picture from the video to each record: the automation names a moment (in milliseconds), Nocarta cuts that frame from the video and adds it to the form record as its own page. The recording itself is never changed.

See the Automation guide for how to write these scripts, including reading your own data with the api statement. Two requirements to remember: the automation needs one of your API tokens bound to it, and this Nocarta address must be on the automation's allowed hosts list.

Programmatic Access

Recordings are also reachable through the API with the media_notes:read and media_notes:write permissions — list and read transcripts, timings and status, upload new recordings, delete and restore. The API never serves the audio or video file itself, only text, timings and status.

Troubleshooting

"A service encountered an error. Our team has been notified and is working on it." — processing failed. The team already has the technical details. Try recording or uploading again.
"Could not access microphone/camera. Please check permissions." — your browser blocked the device; allow access in the browser's site settings.
A long recording seems to fail in chat — the chat waits up to 3 minutes. The recording itself keeps processing in the background and appears with its transcript on the Recordings page.
An empty transcript usually means the recording contains silence — check the player.
Fewer frames than expected on a video is normal: moments that can't be decoded are skipped without failing the recording.