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

# 8. Rough Cut

> Automatically create a rough cut from your raw footage by keeping meaningful sections and removing filler, silence, or irrelevant content.

Automatically create a rough cut from your raw footage by keeping the meaningful sections and removing filler, silence, or irrelevant content — based on your creative intent.

This tile is useful when you want to quickly condense long-form footage into a cleaner baseline edit without manually scrubbing timelines.

## How It Works

The Rough Cut tile analyzes your footage and applies edits based on **your prompt**, not just simple silence trimming. You tell Mosaic what kind of cut you want, and it will:

* Detect meaningful segments
* Remove dead air or unimportant parts
* Preserve emotional, narrative, or informational beats
* Produce a condensed timeline you can build on

## Input Fields

### Creative Intent Prompt (Required)

This prompt defines the *logic* of your cut — **what stays** and **what gets cut**.

Examples:

* "Highlight only the parts where the speaker gives key takeaways."
* "Cut everything except the emotional or dramatic beats."
* "Make an efficient, fast-paced summary version under 90 seconds."

### Quick Suggestions

For users who don't know what to write, Mosaic offers preset buttons for common patterns like storytelling, narrative, tone, and mood-based cuts.

### Multi-Input Behavior

By default, Rough Cut combines all connected input videos into **one cohesive rough cut output**. This is best when you connect many source videos and want Mosaic to choose the strongest moments across the whole set.

Turn off **Combine inputs into one cohesive cut** (`combine_input_renders=false`) when each input should be edited independently. For example, 20 connected videos will produce 20 separate rough-cut outputs instead of one combined timeline.

## Best Use Cases

* Long podcast → short narrative episode
* Interview → highlight reel
* Commentary → tight opinion piece
* Vlog → emotional storyline
* Raw footage → clip-ready baseline

## What Comes Next

Rough Cut produces a clean foundation you can send to:

* [**Clips**](/tiles/clips) — extract short-form highlights
* [**Captions**](/tiles/captions) — add subtitles
* [**Reframe**](/tiles/reframe) — convert to 9:16 for TikTok
* [**Montage**](/tiles/montage) — build a beat-matched sequence
* [**Destination**](/tiles/destination) — export or publish

<AccordionGroup>
  <Accordion title="Example Prompts" icon="pen">
    **Informational:**

    > "Keep only the sections where the speaker explains key concepts or answers questions."

    **Emotional:**

    > "Keep emotional reactions, laughter, suspense, and heartfelt moments."

    **Motivational:**

    > "Extract inspiring sentences and moments where the speaker offers wisdom."

    **Clip Prep:**

    > "Produce a high-signal rough cut that's suitable for extracting shorts."
  </Accordion>

  <Accordion title="Pro Tips" icon="lightbulb">
    * The more specific your prompt, the better the cut
    * Mention **what to keep**, not just what to remove
    * Include mood cues if you care about emotional curation (e.g. "uplifting," "funny," "tense")
    * If you want short-form highlights, combine Rough Cut with [Clips](/tiles/clips)
  </Accordion>
</AccordionGroup>

***

## API Info

<AccordionGroup>
  <Accordion title="Node Params & API Details" icon="code">
    * **Node ID:** `e6098dd4-4a6e-470f-981e-2618c11eee21`

    ### Node params

    | Param                   | Type              | Required | Default  | Notes                                                                                                                                           |
    | ----------------------- | ----------------- | -------- | -------- | ----------------------------------------------------------------------------------------------------------------------------------------------- |
    | `prompt`                | `string`          | Yes      | `""`     | Defines what content to preserve/remove.                                                                                                        |
    | `model_tier`            | `"fast" \| "pro"` | No       | `"fast"` | Controls model quality and speed.                                                                                                               |
    | `combine_input_renders` | `boolean`         | No       | `true`   | Combined output behavior. When `true`, all inputs become one cohesive rough cut. When `false`, each input render gets its own rough cut output. |

    ### Parameter groups

    * **Cut intent (required):** `prompt`
    * **Model quality:** `model_tier`
    * **Combined outputs:** `combine_input_renders`

    ### Behavior notes

    * By default, this tile consolidates multiple inputs into one rough-cut output (`combine_input_renders=true`).
    * Set `combine_input_renders=false` to create one rough-cut output per input render.

    ### Example

    ```json theme={null}
    {
      "prompt": "Keep only concise key takeaways and remove repetitive tangents.",
      "model_tier": "pro",
      "combine_input_renders": true
    }
    ```
  </Accordion>
</AccordionGroup>