Audio production

The Audio room generates speech from your scripts using text-to-speech, handles uploaded voiceovers with timestamp-based splitting, supports sound effect sections, and layers background music under the voice track. Reach it via Audio on the top-chrome stage rail, or from the Script editor with Send to audio / Promote & send.

Script breakdown (project setup)

Pick a script in Review or Final status from the header dropdown. CRAFT parses it into sections and shows you the breakdown before you commit to a voice or mode:

Toggle between AI Voice (TTS) and Upload Recording at the top. The numbered list below the tabs is a live preview of the parsed sections — any row tagged scene is a stage direction that becomes scene context rather than spoken text. Adjust the default pause slider to set the silence between sections (per-section overrides come later in the editor).

Per-section editor (recording & re-record)

Once the project exists the view flips into the 3-column editor. Every parsed beat gets its own row with a waveform, play/regenerate/upload controls, and a delay-after slider. Click any row to focus it in the right rail, where you can edit the text, attach resources, or remove the section entirely:

This is where you iterate. Common moves:

• Regenerate one section — the loop-arrow icon on any row re-runs TTS with the current voice settings. Useful when a single beat comes out with bad pacing.
• Upload one section — replace a single beat with a recorded take without retaking the whole script. Drop a WAV/MP3 on the upload icon; it's stored lossless-first and the waveform updates in place.
• Strip directions — the right-rail shortcut scrubs *(stage direction)* markers from the focused section's text (or use Strip all directions in the left rail to do the whole project).
• Insert SFX between rows — hover the gap between two rows and click +; SFX rows carry an amber sfx chip.
• Tune the pause after any row — the delay-after slider in the right rail overrides the default for just this beat.

Layout

Once a script and an active project are selected, the panel switches to a three-column Audio room layout:

• Left rail — voice header (inline rename), service chip, live stats (sections / ready / pending / duration), primary actions (Generate all · Merge · Strip all directions), collapsible Voice smoothing (TTS-only sliders), collapsible Background bed (upload or MusicGen tabs + volume / fade / mode), collapsible Orphaned (audio files no longer linked to a section).
• Centre — script title + total duration at the top, row-per-beat section list with inline mini-waveform and status chip, insert-section + delay-ms between rows, and a master timeline footer with the merged output's play button and proportional timeline.
• Right rail — focused section detail: text (click Edit for a textarea), any scene context callout, linked resources, and actions (Play · Regenerate · Upload · Record · Strip directions · Remove · delay-after slider).

Prerequisites

• A script in Review or Final status (drafts are blocked).
• A voice configured in Settings — or selected per-project on the first creation screen.

Two modes

TTS (AI voice)

Generate speech from script text using one of four TTS providers:

Provider	Voices	Cost	Quality
Edge TTS	300+ Microsoft Neural voices	Free	Good
ElevenLabs	Premium cloned/generated voices	Paid	Excellent
OpenAI	6 voices	Paid	Very Good
OpenedAI Speech	6 voices (local GPU)	Free	Very Good

TIP

Edge TTS is the default — no API key needed. OpenedAI Speech is also free if you have a GPU. Enable GPU services via Helm (gpu.enabled: true) or docker compose -f docker-compose.dev.yml --profile gpu up -d for local dev.

Upload (recorded voiceover)

1. Create a project in Upload mode.
2. Upload a recorded voiceover (MP3, WAV, OGG, FLAC, M4A).
3. Set start/end timestamps (mm:ss) for each section — the fields auto-format on blur.
4. Click Split All — sections are extracted as lossless WAV files.
5. Re-upload any single section independently if you need to retake just that part.

Creating a project

1. Open the Audio room.
2. Select a script (must be Review or Final).
3. Choose AI Voice (TTS) or Upload mode.
4. If TTS: pick the service and a voice.
5. Click Create Project.

The script is automatically parsed into sections.

Section parsing

The parser tries three strategies in order:

1. Screenplay format — detects stage directions *(...)* and character dialog **Name:** text.
2. Structural markers — splits on ### headings, --- dividers, and **Bold Headers**.
3. Paragraph fallback — splits on double newlines if the above produce only 1 section.

Stage directions

Text in *(stage direction)* markers becomes scene context — not spoken by TTS but shown as an amber callout above the section row that can be toggled on/off with the eye icon.

Sound-effect sections

Insert SFX anywhere in the timeline: hover the gap between two rows, click +, pick Sound Effect. Upload any audio file; SFX rows carry an amber sfx chip and the waveform tracks the imported clip.

Use SFX for transition stingers, intro/outro jingles, or ambient breaks.

Background bed

The Background bed section in the left rail controls the ambient layer:

• Upload — pick a file or drop it in. Configure volume, fade in/out, and mode (Loop / Trim to voice / Play once).
• Generate (tab, if MusicGen GPU service is running) — describe the music, set duration, generate.

Settings are applied during Merge.

Standalone music generation

A dedicated Music Generator panel lives at activeView: 'audio-music' (reachable from the command palette or the resource sub-nav when applicable). It creates tracks without needing a project — useful for quickly auditioning ideas.

Generating audio

• Generate all — left-rail button, processes every pending section sequentially.
• Regenerate (right rail / row icon) — re-generates a single section with current voice settings.

Each row shows its status: pending / generating / ready / error. Errors get an inline message plus a red border.

Voice controls

In the left rail's Voice smoothing section:

• Speed — playback speed multiplier (Edge TTS, OpenAI, OpenedAI Speech)
• Stability (ElevenLabs) — lower = more expressive, higher = more consistent
• Similarity boost (ElevenLabs) — how closely to match the original voice

Linking resources

In the right-rail section detail, click Attach. The picker shows downloaded resources from the channel library. Linked resources appear as chips on the row — useful for tracking which B-roll goes with which section when later assembling video.

Merging

Click Merge all in the left rail to combine every ready section into a single audio file:

• Silence gaps between sections come from each row's delayAfter slider.
• Mixed formats (WAV + MP3) are normalised automatically.
• If a background track is configured, it's layered under the voice with fade in/out.

Once merged, the master timeline footer at the bottom of the centre column shows the play button, total duration, and a Finalize & create episode button that promotes the project into an Episode.

Text preprocessing

Edge TTS doesn't support SSML, so text is automatically preprocessed:

• Dashes become commas (natural pauses)
• Ellipses become periods (longer pauses)
• Units spelled out: °C → "degrees C"
• Abbreviations expanded: e.g. → "for example"