Sessions and Tracks
The session graph, undo and redo, how audio is loaded onto tracks and mixed with lane controls, and recording and punching in.
On this page3 topics
Sessions and the Session Graph
Every edit creates a session node, a snapshot of the session. Nodes link to their parent in a directed acyclic graph (DAG), so nothing is destructive: your original audio files are never modified and you can return to any earlier state.
- Graph view — the Graph tab, beside Timeline and Transcript, draws all session nodes with their label, the tool that made them and their age.
- Preview a version — click a node in the graph to render that version's mix and play it. The node gets a dashed outline and your place in the history does not move: the next edit or undo starts from where you were.
- Revert to any node — right-click a node in the graph and choose "Set as head" to jump back to that state. Undo brings you back.
- Fork — branch from any node to try another direction.
- Undo and redo —
Ctrl/Cmd+ZandCtrl/Cmd+Shift+Zretrace the steps you took, the agent's edits included. The path is saved with the project, so it survives a restart. - Compact history — ask
this project is huge — reclaim some diskand the agent reports what it would remove, then, if you confirm, deletes old undo steps for good, with the audio the project made for them. A file you loaded is never deleted. The most recent steps stay, so ordinary undo keeps working. - One writer per project folder — opening a project is refused while a batch run is writing its own project into that folder, and while the folder was replaced at the same path (restored from a backup, or trashed and made again) and the assistant is still working in the old one. Wait for the turn to end, then open it again. A batch run is likewise refused, and starts nothing, if one of the projects it would write is open.
Tracks
A session has one or more tracks, each a sequence of clips. A clip describes a region of a source file on disk, not a copy of it.
- Load audio — drag a file onto the timeline, click “Open Audio”, or type
load /path/to/file.mp3. Formats: WAV, MP3, FLAC, Ogg Vorbis, Opus (mono or stereo) and AIFF, and M4A files with AAC audio on macOS and Windows, where the system decodes it. - Multi-track sessions — load several files or ask:
add an empty track called drums. - Lane controls — each track's lane has its own gain slider, pan slider, mute and solo buttons, so you can mix by hand. Double-click a pan slider to centre it. A mute silences a track without removing it; a solo plays only the soloed tracks. One drag or click is one history step, and a control you have just set stays put until the session confirms it.
Recording and Punching In
Record in the header captures the default microphone to a new track; Stop adds it to the session. The clock button next to it records on a timer, unattended.
Punch replaces one region of a track with a new take, so a misread line is fixed without re-recording the rest. Select the region on the timeline, press Punch, choose the track and the lead-in (two seconds by default), and press Start. You hear the whole mix through the speakers from the lead-in on, with the track you are replacing silent inside the region and heard again after it; the lead-in and the tail are what you play along with. The new take lands in the session as one history step, the length of the region, and everything after it stays where it was. Cancel on the strip above the timeline throws the take away and leaves no step.
A take comes back to the microphone later than the sound that cued it left the program: output and input buffers, the converters, the speaker and the air. Cut at the region's own time it would land late by that much, every time, so a punch needs the delay for your speakers and microphone, and it will not guess one. In Settings → Audio either press Measure, which plays test tones through the speakers and listens for them (put the speaker close to the microphone, or hold the headphones against it, in a quiet room), or type the delay in milliseconds if you can't make noise in the room or found it another way, and press Use this value. Either counts, the newest one is kept, and it is kept for that pair of speakers and microphone: a headset or a different sample rate is another pair. Until one is kept the Punch popover says “Not calibrated” and Start stays off. A delay that an earlier version of edytlab kept, measured or typed, is not used by version 0.9.4 or later, because the audio library changed and with it the delay; measure or type it once more.
- Headphones keep the mix out of the microphone. The delay measured with the speaker next to the microphone is right for headphones on the same device; a Bluetooth headset is a different device and has to be measured or typed on its own.
- The take is put in the track's channels by what the microphone delivered, not by position: a microphone on one input of a two-input interface, with the other silent, comes out on both channels of a stereo track (and at full level in a mono one), not on the left alone. If both inputs carry sound, a mono track gets the louder one and a stereo track gets both.
- While a punch-in runs, Record, the timer and a measurement are off, and the other way round: they share the microphone and the speakers.
- If audio was lost during the take, or the project changed while you played, the take is not applied (in the second case it is kept, and the message says where).
- The speakers and the microphone are clocked separately, so a long take drifts by a few milliseconds over a minute. That is nothing for a line; for a verse, punch in sections.