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

# TTS, Batch & Prompts

> Text-to-speech voices, batch media processing, and LLM prompt overrides

# TTS, Batch & Prompts

Three utility command groups: `omni tts` for voice synthesis, `omni batch` for bulk media processing, and `omni prompts` for customizing LLM prompt templates.

## TTS Commands

### `omni tts voices`

List available text-to-speech voices.

```bash theme={"dark"}
omni tts voices
```

TTS is also available inline via `omni send --tts "text"` — see [Messaging](/omni/cli/messaging) for details.

***

## Batch Commands

The `omni batch` command group manages bulk media processing jobs — transcribing audio, describing images, and extracting document content at scale.

### `omni batch create`

Create a batch processing job.

```bash theme={"dark"}
omni batch create [options]
```

| Option                    | Description                                                                |
| ------------------------- | -------------------------------------------------------------------------- |
| `--instance <id>`         | Instance ID                                                                |
| `--type <type>`           | Job type: `targeted_chat_sync`, `time_based_batch`, or `media_redownload`  |
| `--chat <id>`             | Chat ID (required for `targeted_chat_sync`)                                |
| `--days <n>`              | Days to look back (required for `time_based_batch` and `media_redownload`) |
| `--limit <n>`             | Max items to process                                                       |
| `--content-types <types>` | Content types: `audio`, `image`, `video`, `document` (comma-separated)     |
| `--force`                 | Re-process items that already have content                                 |
| `--delay-min <ms>`        | Min random delay between items in ms (default: 1000)                       |
| `--delay-max <ms>`        | Max random delay between items in ms (default: 3000)                       |
| `--no-confirm`            | Skip confirmation prompt                                                   |

### Job Types

| Type                 | Description                                       |
| -------------------- | ------------------------------------------------- |
| `targeted_chat_sync` | Process media from a specific chat                |
| `time_based_batch`   | Process media from all chats within a time window |
| `media_redownload`   | Re-download and reprocess media files             |

### Examples

```bash theme={"dark"}
# Transcribe all audio from a chat
omni batch create \
  --instance <id> \
  --type targeted_chat_sync \
  --chat <chat-id> \
  --content-types audio

# Process last 7 days of images and documents
omni batch create \
  --instance <id> \
  --type time_based_batch \
  --days 7 \
  --content-types image,document

# Force reprocess with no confirmation
omni batch create \
  --instance <id> \
  --type time_based_batch \
  --days 30 \
  --force \
  --no-confirm
```

### `omni batch list`

List batch jobs.

```bash theme={"dark"}
omni batch list [options]
```

| Option              | Description                                                                                  |
| ------------------- | -------------------------------------------------------------------------------------------- |
| `--instance <id>`   | Filter by instance ID                                                                        |
| `--status <status>` | Filter by status: `pending`, `running`, `completed`, `failed`, `cancelled` (comma-separated) |
| `--type <type>`     | Filter by job type                                                                           |
| `--limit <n>`       | Max results                                                                                  |

```bash theme={"dark"}
omni batch list
omni batch list --status running,pending
omni batch list --instance <id> --type time_based_batch
```

### `omni batch status`

Get job status with optional live updates.

```bash theme={"dark"}
omni batch status <job-id> [options]
```

| Option            | Description                         |
| ----------------- | ----------------------------------- |
| `--watch`         | Watch for updates                   |
| `--interval <ms>` | Poll interval in ms (default: 2000) |

```bash theme={"dark"}
omni batch status <job-id>
omni batch status <job-id> --watch
```

### `omni batch cancel`

Cancel a running job.

```bash theme={"dark"}
omni batch cancel <job-id>
```

### `omni batch estimate`

Estimate job scope and cost without creating the job. Accepts the same options as `create`.

```bash theme={"dark"}
omni batch estimate [options]
```

| Option                    | Description                        |
| ------------------------- | ---------------------------------- |
| `--instance <id>`         | Instance ID                        |
| `--type <type>`           | Job type                           |
| `--chat <id>`             | Chat ID (for `targeted_chat_sync`) |
| `--days <n>`              | Days to look back                  |
| `--limit <n>`             | Max items                          |
| `--content-types <types>` | Content types to include           |

```bash theme={"dark"}
# Check how many items before committing
omni batch estimate \
  --instance <id> \
  --type time_based_batch \
  --days 30 \
  --content-types audio,image
```

***

## Prompt Commands

The `omni prompts` command group manages LLM prompt templates used for media description (images, videos, documents) and gating decisions. Override defaults to customize how Omni processes media.

### `omni prompts list`

List all prompt settings with their override status.

```bash theme={"dark"}
omni prompts list
```

### `omni prompts get`

Show the current prompt for a specific type.

```bash theme={"dark"}
omni prompts get <name>
```

| Name       | Description                            |
| ---------- | -------------------------------------- |
| `image`    | Prompt for describing images           |
| `video`    | Prompt for describing videos           |
| `document` | Prompt for extracting document content |
| `gate`     | Prompt for gating decisions            |

```bash theme={"dark"}
omni prompts get image
omni prompts get gate
```

### `omni prompts set`

Set a prompt override. Reads from stdin if no value is provided (for multiline prompts).

```bash theme={"dark"}
omni prompts set <name> [value] [options]
```

| Option              | Description       |
| ------------------- | ----------------- |
| `--reason <reason>` | Reason for change |

```bash theme={"dark"}
# Inline override
omni prompts set image "Describe this image in detail, including text and objects" --reason "More detailed descriptions"

# Multiline via stdin
cat <<'EOF' | omni prompts set document --reason "Custom extraction"
Extract all text from this document. Include:
- Headers and titles
- Body text
- Table contents as markdown
EOF
```

### `omni prompts reset`

Clear a prompt override, reverting to the code default.

```bash theme={"dark"}
omni prompts reset <name>
```

```bash theme={"dark"}
omni prompts reset image
```

<Warning>
  `omni prompts set` overrides the global default for **every** instance, agent, and incoming media of that type until reset. Test on a non-production instance first — a bad transcription prompt can silently corrupt downstream agent context for hours before anyone notices.
</Warning>

<Tip>
  Use `--reason` on every `prompts set` so the next operator can tell why an override exists. The reason is surfaced in `omni prompts list`.
</Tip>

## See also

<CardGroup cols={2}>
  <Card title="Messaging" icon="message" href="/omni/cli/messaging">
    Inline TTS via `omni send --tts` and the `omni speak` verb.
  </Card>

  <Card title="Media architecture" icon="image" href="/omni/architecture/media">
    How media flows through transcription, description, and extraction.
  </Card>
</CardGroup>