Editor

Home · Manual index

Editor tab

The nav label is Editor. Internally it is the jobs tab (data-tab="jobs", tab-jobs.js). It is a three-pane single-track audio editor over files the job system already produced. Renders are non-destructive: they write a new file under <job_dir>/edits/ and leave the original FINAL / chunk intact.

Layout

The blue area marks a selection. The three displayed effects are full-track effects; a selection only limits an operation added with a range button.

[ Library | Editor (waveform + transport) | Effects rack ]

The canvas is zero-sized while the tab is hidden; switching to Editor re-measures on the next frame. If the waveform looks empty after a generate, click Editor once and wait a tick, or click Refresh.

Library (left)

Header Library plus Refresh (GET /api/jobs via App.pollJobs()).

Search box: Search jobs / text... filters on text_preview, model, and job_id.

Empty: No jobs yet — generate something on the Testing tab. or No matches.

Each job is a foldable row:

• Caret, a title (first 60 characters of the preview, or model · first-8-of-uuid), and a status badge (completed / running / partial / failed).
• Click the header to expand. Right-click for a job context menu.

When expanded:

• FINAL — the assembled file (kind: final, job_id) if final_file is set.
• CHK N — each completed chunk WAV (chunk_000.wav, …). Pending chunks show N chunk(s) not yet generated.
• EDIT — previous renders from GET /api/audio/edits/{job_id}. Each edit has a ✕ to delete (DELETE /api/audio/edits/{job_id}/{filename}).

Click a FINAL / CHK / EDIT row to load it as the source. The active row is highlighted.

Center: waveform and transport

Top toolbar:

Control	Role
Source name	e.g. job preview · final
Source meta	duration / format
Save As...	Choose an edit name and format; saves under the job's edits/ directory
Download	Fetches the current source audio through the authenticated blob helper

The canvas shows min/max peaks from POST /api/audio/peaks (buckets default 1024). Overlay: selection rectangle and playhead. Empty text: Pick a track from the library to start editing. Loading text: Loading....

Transport:

Control	Role
▶	Play / pause the current source (not a live preview of unrendered effects)
■	Stop, playhead back
Time	00:00.00 / 00:00.00
Selection time	start–end when a range is selected
Cut	Range op cut
Trim to Selection	Range op trim
Silence	Range op silence
Fade In	Range op fade_in
Fade Out	Range op fade_out
Zoom	1.0× to 40×

Range buttons stay disabled until you drag a selection on the waveform. Range ops are queued in the effects chain (they run before full-track effects on Render). Rendering writes a new edit; the source file is preserved.

Click-drag on the canvas sets start_sec / end_sec. Zoom slider changes _zoom; you can pan via view offset when zoomed in.

Effects rack (right)

Header Effects plus Clear (empties the chain).

+ Add Effect opens a menu grouped by category (from FX_DEFS in tab-jobs.js):

Category	Effect label	Type id	Parameters (GUI min/max/default)
Volume	Gain	gain	Gain -24..24 dB, default 0
Time / Pitch	Speed (Tempo)	tempo	Factor 0.5..2.0×, default 1.0 (rubberband)
Time / Pitch	Pitch Shift	pitch	Semitones -12..12, default 0
EQ	High-pass	eq_highpass	Cutoff 20–1000 Hz, default 80
EQ	Low-pass	eq_lowpass	Cutoff 1000–20000 Hz, default 12000
EQ	Bass Shelf	eq_shelf_low	Frequency 50–500 Hz default 200; Gain -12..12 dB
EQ	Treble Shelf	eq_shelf_high	Frequency 1000–12000 Hz default 4000; Gain -12..12 dB
Space	Reverb	reverb	Room 0–1 default 0.5; Damping 0–1 default 0.5; Wet 0–1 default 0.3
Space	Echo / Delay	echo	Delay 0.05–1 s default 0.25; Feedback 0–0.95 default 0.4; Wet 0–1 default 0.4
Dynamics	Compressor	compressor	Threshold -40..0 dB default -20; Ratio 1–20 default 4; Attack 1–200 ms default 10; Release 10–1000 ms default 150; Makeup 0–24 dB
Dynamics	Noise Gate	noise_gate	Threshold -80..-20 dB default -50; Attack 1–50 ms default 5; Release 10–500 ms default 100
Restoration	De-reverb	de_reverb	Strength 0–1 default 0.5
Restoration	De-esser	de_ess	Strength 0–1 default 0.5
Mastering	LUFS Normalize	normalize_lufs	Target -36..-6 LUFS default -16
Mastering	Peak Normalize	normalize_peak	Target 0.5–1 default 0.95

Each chain item can be enabled/disabled and has its sliders. Order is the render order after range ops.

Footer: format select wav / flac / ogg / mp3 and primary Render.

Render calls POST /api/audio/render with:

{
  "source": {"kind": "final", "job_id": "..."},
  "edits": [{"type": "gain", "params": {"db": 3}}],
  "output_format": "wav",
  "overwrite": false
}

Job-bound sources default to an auto-versioned name under <job_dir>/edits/ when you do not Save As. A path source (CLI/API only from outside the job library) requires output_path under voices, output, or projects_output.

A missing effect dependency or an invalid edit makes Render fail with an explicit error. Install the dependency and retry; the server does not publish a file with silently skipped effects. A chain that removes all audio is rejected. Stereo sources retain their channels.

After Render, the new EDIT appears in the library. You can load it and stack another chain. Changes made to the chain or source while rendering are preserved. Range selections stay anchored to the waveform through earlier cuts, trims, padding, and tempo changes.

Save As and Download

Save As... opens a dialog with Name and Format. Enter a simple filename such as studio_demo_master and choose WAV, FLAC, OGG, or MP3, then click Save. This renders the current effects into a named file under the job's edits/ directory. You can omit the extension. Absolute output paths are available through the CLI/API, not this dialog; API output_name and output_path are mutually exclusive.

Download pulls the current audio through the token header so you can save a copy on the Windows side without browsing \\wsl$.

Save As renders the queued effects into a named edit. Play and Download use the currently loaded source, so render first to hear or download your changes.

After a successful save, the EDIT is loaded and the applied effects chain clears. The original FINAL and CHK rows remain available.

What the Editor tab is not

It is not a multi-track DAW. It is not where you change TTS text (use tts.cmd jobs edit-chunk / PUT /api/jobs/{id}/chunks/{idx} and rerun). It does not generate SRT (see the SRT page). It lists jobs but does not delete them (use tts.cmd jobs delete or POST /api/jobs/delete; running jobs are refused).

Related pages

• Audio editor usage (full effect contract + CLI)
• Jobs and projects
• Testing tab