A recording is a source like any other: you add an audio or video file of a meeting, it is transcribed, and the draft is built from what was said. The transcript keeps each speaker and the time of each line, so a draft can use who said what and when.
If your meeting tool already gave you a transcript file, add that file instead. It costs no credits to transcribe. See Source materials.
Before you start
- Recordings are available on Core, Team and Enterprise. On other plans the source picker names the plans that have them.
- Recording transcription must be on for the workspace. It is on unless an owner or admin turned it off. See "The recording transcription switch" below.
- A recording can be up to 3 hours long and up to 2 GB.
- These file types are accepted: .mp3, .m4a, .wav, .aac, .ogg, .opus, .flac, .webm, .mp4, .mov, .mkv.
- Transcribing costs 5 credits for each minute of the recording, rounded up to the whole minute. The price is shown on the row while the recording uploads. Check the workspace's balance in Settings → Credits & Billing.
- Any signed-in member of the workspace can add a recording. There is no owner or admin requirement.
- Use a recording you have the right to share. See "What is sent to the transcription service" below.
Steps
- Open a place where you add sources: the source picker in the Studio create dialog or on the full create forms, the Knowledge library in Studio → Knowledge, or the attach menu in Ask devlin.
- Drop the recording on the picker, or choose it from your computer. The row shows "Checking the recording…" while its length is read in your browser. Nothing is uploaded yet.
- The upload then starts on its own. The row shows "Uploading" with a percentage and "Keep this page open." Under the file name, a line gives the length and the price: "Recording," followed by the length, then "Transcribing uses" followed by the number of credits, which is 5 credits for each started minute. The line below it says where the recording will be kept: "Kept for this build only, for 7 days." or "Saved to your library."
- Wait while it uploads. The upload goes from your browser straight to storage. If the connection drops, it resumes on its own. Closing the page stops it, and your browser asks before it closes. Closing the create dialog, or going to another page in devlin.ai, also stops it. Select Cancel to stop the upload yourself. No credits are used for an upload that is stopped.
- When the row shows "Transcribing…" and "You can leave this page.", the upload is done. Transcribing carries on without the page. There is no percentage or time estimate for this step.
- When the row shows "Ready" with the length and the number of speakers, the recording can be used like any other source.
In Ask devlin the same steps happen on the attachment chip: the upload starts on its own, and the chip shows the upload's progress with the length and the price. A message sent while the recording is uploading or being transcribed waits for it and is sent when the recording is ready.
Result
The recording becomes a source with a transcript. Drafts, searches and summaries read the transcript the way they read any document.
- Each line of the transcript starts with the speaker and the time, for example "Speaker 2 [00:06:41]:". Speakers are numbered in the order they first speak. Names are not recognized.
- The transcript has a heading for each stretch of the meeting, so a long meeting can be read in parts.
- The original recording is kept with the source. It is not played back anywhere in the app.
- A long transcript is summarized the same way a long document is. Source materials explains summaries and what they cost.
What it costs
- Transcribing costs 5 credits for each minute, rounded up to the whole minute.
- The credits are set aside when the upload starts, before any of the recording is sent. If the workspace does not have enough, the row says how many are needed and nothing is uploaded.
- If a recording in your workspace once turned out much longer than its file reported, later recordings in the workspace have the most their file size allows set aside, up to 3 hours of transcription, instead of the length shown. The amount needed to start can then be more than the price, and once the upload starts the row names it after the price: "are set aside until it is done." The charge is still for the length that was transcribed, and the rest is given back.
- The charge is made when the transcript arrives, for the length that was transcribed. It is never more than the amount set aside. A recording that runs longer than its file reported is charged the amount set aside.
- No credits are used when the upload is cancelled or does not finish, or when the recording could not be transcribed. The amount set aside is given back.
- A recording in which no speech is found was still transcribed, and is charged.
- A recording you delete after it has been sent to be transcribed is still charged the amount set aside, because the transcription had already started. A recording deleted while it is still waiting for its turn (its row also shows "Transcribing…") is not charged, and the amount set aside is given back.
Where recordings are kept
- A recording saved to the library is kept until you delete the source. Deleting the source deletes the original recording, the transcript and everything read from it.
- A recording added for one build only is removed 7 days later, like any build-only source.
- An upload that is not finished within 24 hours is removed and its credits are given back.
- Recordings have their own storage allowance for the workspace: 10 GB on Core, and 50 GB on Team and Enterprise. Recordings do not count against the Knowledge library's storage or against hosting. A recording saved to the library counts as one library document (its transcript). A recording kept for this build only does not count against the library's limits.
- A recording that fails after it was sent to be transcribed stays, with its reason, until you delete it, and counts against the allowance until then. A recording refused right after its upload finishes is removed on its own, and its credits are given back. Removing it from a create form's list or a chat chip does not delete it: delete a failed library recording in Studio → Knowledge. A build-only one is removed 7 days later.
What is sent to the transcription service
To make the transcript, devlin.ai gives its speech-to-text provider a short-lived link to the recording. The provider reads the recording through that link and returns the transcript. devlin.ai then stores the transcript and asks the provider to delete its copy.
The recording transcription switch
Each workspace has a switch that decides whether its recordings may be sent to the transcription service. It is on by default.
- Open Settings → Data & privacy as an owner or admin of the workspace.
- In the Recording transcription card, select Turn off, or Turn on to allow it again.
While it is on, the card reads "On: meeting recordings added as sources are transcribed." While it is off, it reads "Off: no recording from this workspace is sent to the transcription service." Members can open the page and see whether the switch is on, with "This setting is managed by your workspace owner and admins." They cannot change it.
While the switch is off:
- A recording cannot be added. Dropping one on a source picker says "Recording transcription is turned off for this workspace. An owner or admin can turn it on in Settings, Data & privacy." Nothing is uploaded and no credits are set aside.
- A recording that was uploaded and is still waiting for its turn to be transcribed is stopped, and the credits set aside for it are given back. It shows "This recording could not be transcribed. No credits were used. Try again, or upload a transcript file."
- A recording already at the transcription service finishes and is charged as usual.
- Transcript files can still be added. See Source materials.
- Recordings and transcripts you already have are kept.
Each change of the switch is recorded in the activity log of an organization workspace.
Limits
- Length: 3 hours for one recording.
- Size: 2 GB for one recording.
- Uploads started by one person: 10 in an hour. A start refused because the file is not a recording, is too large or too long, there is no room, or there are not enough credits does not count. A start refused by the workspace's daily upload limit does count.
- A workspace can upload up to 20 GB of recordings a day on Core, and 100 GB on Team and Enterprise. Deleting a recording does not give that day's amount back.
- A recording that is still not transcribed after 2 hours is stopped, and no credits are used.
What gets refused
- A recording that is too long or too large is refused before anything is uploaded, with the limit in the message. Trim it, split it into parts, or export it at a lower quality or as audio only.
- "This file could not be read as a recording." means your browser could not read the file's length. Export the recording again from your meeting tool, or convert it to one of the accepted file types.
- "This file's length could not be confirmed. Export it again and retry." means the length the file reports does not fit its size. Export it again.
- "Your workspace's recordings storage is full. Delete a recording to make room." means the workspace's recordings allowance is used up.
- "Your workspace has reached today's recording upload limit. Try again tomorrow." and "You have started too many recording uploads. Try again in an hour." are the upload limits above.
- "Recording transcription is not available right now." means transcription is switched off for the moment. Nothing was uploaded. Try again later, or add a transcript file.
When a recording fails
- "The uploaded file does not match what was declared." The file that reached storage was not the size the browser reported. The recording has already been removed and no credits were used. Add the recording again.
- "This recording could not be processed. Upload it again." The recording was refused when it was sent to be transcribed. It has already been removed and no credits were used. Add it again, or add a transcript file.
- "No speech was found in this recording." The file had no speech that could be transcribed. Check that the recording has an audio track.
- "This recording could not be transcribed. No credits were used. Try again, or upload a transcript file." Delete the failed recording from the library, then add the recording again.
- "Transcribing this recording took too long and was stopped. No credits were used. Try again." Delete the failed recording from the library, then add the recording again. This also happens when a recording has waited too long for its turn to be transcribed.
A plan without recordings
On Free and Starter, the source picker says "Meeting recordings are available on the Core, Team and Enterprise plans." under its list of file types. A recording dropped on it says "Recordings are available on the Core, Team and Enterprise plans." and nothing is uploaded. The Recording transcription card in Settings → Data & privacy is faded, with "Not on this plan" and a link to the plans.