> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ozu.studio/llms.txt
> Use this file to discover all available pages before exploring further.

# Audio Node

> Generate voiceover and dialogue lines from your script, audition takes, and sync approved takes to the Timeline.

The **Audio Node** is where your script becomes voice. You type dialogue (or pull it from the script), pick a voice, generate takes, and approve the ones you want. Approved takes flow into the Timeline's Dialogue track automatically.

## What the Audio Node is for

One Audio Node serves one purpose: **generating and approving voiceover for dialogue lines**. Each line of dialogue in your scene is its own entry — generated independently, auditioned independently, approved independently.

This mirrors how dialogue works in real production: every line gets its own performance, and you pick the take you want before the cut.

## Adding dialogue

Two ways to get dialogue into the Audio Node:

<Columns cols={2}>
  <Card title="From the script">
    If your Script Node has analyzed a screenplay, dialogue lines are already extracted as assets. The Audio Node picks them up automatically — each line appears as a row, tagged with the beat it belongs to.
  </Card>

  <Card title="Typed directly">
    For narration, voiceover, or lines you didn't include in the script, type them in. Each line you add becomes a row you can generate takes for.
  </Card>
</Columns>

## Picking a voice

Each dialogue line has a **voice** setting. The Audio Node gives you a library of voices covering:

* **Male / female / neutral**
* **Age range** — child, young adult, middle-aged, elder
* **Accent** — neutral, British, American regional, various
* **Tone** — warm, authoritative, anxious, laid-back, dramatic

Pick a voice per character — for example, one voice for every line Kael speaks, another for Mira. The Audio Node remembers character-voice assignments so you don't have to set the voice line by line.

<Tip>
  **Audition before committing.** Each voice has a short preview sample. Play a few before assigning one to a character — AI voices carry a lot of personality, and the wrong fit is obvious three lines in.
</Tip>

## Generating takes

Click **Generate** on a dialogue line and Ozu produces a **take** — one performance of that line in the chosen voice.

You can generate multiple takes per line. Each take is a separate audio clip you can play back, compare, and choose from.

1. **Open a dialogue line** in the Audio Node.

2. **Set the voice** (or inherit the character's default).

3. **Click Generate.** A take renders and becomes playable.

4. **Play the take.** If it works, approve it. If not, generate another.

5. **Repeat until you have the take you want.**

## Approving and rejecting takes

Every generated take is in one of three states:

* **Pending** — generated but not yet reviewed
* **Approved** — the take you want to use; flows to the Timeline
* **Rejected** — not going to use this one; hidden from the Timeline

Only **one approved take per line** can exist. Approving a new take auto-rejects the previous one. This keeps the Timeline clean — you always have exactly one voice performance per line.

<Note>
  **Rejected takes aren't deleted.** They stay in the take history for that line. You can un-reject and re-approve later if you change your mind.
</Note>

## Syncing to the Timeline

Approved dialogue takes flow automatically into the Timeline's **Dialogue track**, aligned to the beat they belong to. The flow is:

1. **Audio Node generates a take.**

2. **You approve it.**

3. **The Timeline's Dialogue track updates** — a clip appears at the beat's position on the track.

4. **Open the Timeline** to verify sync. Scrub to the beat and play.

If the dialogue line's beat position changes (you edited the script, reordered beats), the Timeline clip follows. The approval is sticky to the line, not the timecode.

## Editing a line's timing

The initial timecode of a dialogue clip is inherited from its beat. To retime — delay a line, overlap it with action, sync to a specific moment — drag the clip on the Timeline's Dialogue track.

For tighter edits (trim the head, fade the tail), use the Timeline's clip controls. The Audio Node itself doesn't handle timeline-level editing.

## Line-level vs. project-level voice

Two layers of voice setting:

* **Project default** — if a line has no specific voice set, it uses a fallback you've configured for the project.
* **Character default** — the voice assigned to each named character (Kael, Mira, etc.). Inherits down to every line that character speaks.
* **Line override** — an individual line can override the character's voice. Useful for flashbacks, alternate performances, or one-off effects.

## Related

<Columns cols={2}>
  <Card title="Song Node" href="/user/audio-and-music/song-node">
    The other audio node — music and score, not dialogue.
  </Card>

  <Card title="Tracks" href="/user/timeline/tracks">
    Where approved takes land — the Timeline's Dialogue track.
  </Card>

  <Card title="Captions and Markers" href="/user/timeline/captions-and-markers">
    Generate captions automatically from approved dialogue takes.
  </Card>

  <Card title="Script and Analysis" href="/user/script/writing-your-script">
    Where dialogue originates — the Script Node's analysis of your screenplay.
  </Card>
</Columns>

## FAQ

**Do I pay credits per take, or per approved take?**
Credits are held when a take starts generating and settled when it succeeds. If a generation fails, no charge. Every take counts — so compare carefully before generating a third or fourth variant.

**Can I upload my own voice recording instead of generating?**
Yes. The Audio Node accepts uploaded audio files per line. Drag a file onto a dialogue row and it becomes the approved take for that line — no AI generation involved.

**Why doesn't my approved take show up on the Timeline?**
Check three things: the Audio Node is connected to the Timeline; the dialogue line has a valid `beatId`; the line's beat exists in the storyboard. If any one is missing, the sync skips the line.

**Can I mix AI-generated and uploaded takes in one project?**
Yes. Some characters' dialogue is generated, some is you recording on a phone — the Audio Node treats them identically once approved.

**What happens to approved takes if I re-run script analysis?**
Approved takes persist. The Audio Node matches by dialogue line ID, which is stable across re-analysis as long as the line's text doesn't change meaningfully. If you rewrite a line, you'll need a new take.

**Can I generate all my dialogue at once?**
The Audio Node supports batch generation — select multiple lines and hit generate. It's the fastest way to draft a full scene's voiceover, but listen through before approving; the first pass won't always match the performance you want.