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

# Write prompts, references and commands

> How to describe edits, reference clips and transcript passages with @, run / commands, attach files, use voice input and set generation options

The composer is where you direct the edit. Plain language works for most requests; `@` references, `/` commands, attachments and generation settings make your prompts precise. Read this after [Meet GoFrame Editor Agent](/editor/agent).

## Describe the edit

Say what you want and where. Name the clip, track or moment when it matters, and state the outcome rather than the steps.

| Prompt                                                                         | What the agent does                                                                 |
| ------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------- |
| "Remove filler words and pauses longer than a second from the interview."      | Edits the timeline through the transcript, then shows a **Timeline changes** card.  |
| "Cut this into three vertical shorts, 30 seconds each, with karaoke captions." | Finds highlights, builds three clips and applies a Karaoke caption preset.          |
| "Add a lower third with the speaker's name at 0:05, our brand blue."           | Generates a motion graphic and places it at that time.                              |
| "Generate a 6-second aerial shot of a coastline at sunrise for the intro."     | Announces the cost, runs a video generation, and puts the result in the Media pool. |
| "Why does the second scene feel slow?" (in **Ask** mode)                       | Answers without changing anything.                                                  |

<Tip>
  Give constraints up front: target length, aspect ratio, tone, and what must not change. The agent reads your rules and memory too; see [Agent rules, memory and voice](/editor/agent-settings).
</Tip>

Under the text field sit the chat mode dropdown, **Plan**, **Generation settings**, **Select AI models** (✦), **Attach files**, **Voice input** and **Send message**. The rest of this page covers each of them.

## Reference clips, assets and transcript with @

Type `@` to open the **Assets & Clips** palette. It lists the **Clips** on the timeline and the assets in the Media pool. Pick an entry and it is inserted into your message as a reference the agent can resolve exactly. Selecting a clip on the timeline also adds a context chip with its asset name above the composer, so the agent knows what you are looking at.

* `@` a clip when you mean one specific item on the timeline ("trim @intro-take-2 to 8 seconds").
* `@` an asset to use a file that is not on the timeline yet ("put @logo.png in the corner for the whole video").
* To point the agent at spoken content, use **Add to chat** in the Transcript panel (below).

<Frame>
  <img src="https://mintcdn.com/goframe-fb816310/nCYzUacADHcqMX4R/images/editor/prompting-mentions.png?fit=max&auto=format&n=nCYzUacADHcqMX4R&q=85&s=0450ffbeb70b8723c3782c608779471d" alt="The Assets & Clips palette opened with @ showing clip entries" width="810" height="820" data-path="images/editor/prompting-mentions.png" data-path="images/editor/prompting-mentions.png" />
</Frame>

### Add a passage from the Transcript panel

In the **Transcript** panel, select a passage and click **Add to chat**. The text arrives in the composer with its timing, so you can say "make this the opening line" or "cut everything between these two sentences". See [Edit video by editing text](/editor/transcript-editing).

## Run commands and skills with /

Type `/` to open the command palette. It has two sections: **Commands**, each tagged **RUN** or **SEND**, and **Skills**, each tagged with what it does (**GENERATE**, **EDIT** and so on) and a hint for the argument it takes. Skills are the agent's specialized abilities (see [Skills and workflows](/editor/skills)).

| Command       | What it does                                                                              |
| ------------- | ----------------------------------------------------------------------------------------- |
| `/compact`    | **Compact context**: summarize the session and free up space.                             |
| `/undo`       | **Undo last change**: revert the most recent agent-made edit to the project.              |
| `/plan`       | **Enable Plan mode**: the next turn runs read-only and produces a plan before any change. |
| `/clear`      | **Clear chat**: start a new chat and keep the project state.                              |
| `/explain`    | **Explain (dry run)**: get a read-only plan for the next step without executing anything. |
| `/reasoning`  | **Toggle reasoning mode**: switch between Agent and Reasoning.                            |
| `/model`      | **Change model**: open the model picker. See [AI models](/editor/ai-models).              |
| `/director`   | **Run director skill**: start the full 6-stage creative pipeline.                         |
| `/storyboard` | **Run storyboard-preview**: generate a single-page storyboard of the current project.     |
| `/polish`     | **Run polish skill**: final cleanup pass on audio, cuts and fades.                        |

Any skill can be called the same way, for example `/image-gen a neon-lit rainy street at night`.

<Frame>
  <img src="https://mintcdn.com/goframe-fb816310/nCYzUacADHcqMX4R/images/editor/prompting-palette.png?fit=max&auto=format&n=nCYzUacADHcqMX4R&q=85&s=a16ee0279cb683bbae16ad04d5578308" alt="Slash command palette open with Commands and Skills sections" width="810" height="980" data-path="images/editor/prompting-palette.png" data-path="images/editor/prompting-palette.png" />
</Frame>

## Attach files

Drag files onto the chat or click **Attach files** in the composer to add context: a script, a brief, a brand PDF, a shot list. Attached files appear as chips above the text (for example **artemis-brief.pdf · PDF · 298 B**, or a thumbnail for an image).

* Up to **10 files** per message, **25 MB** each.
* Supported: images, PDF, text (`.md`, `.txt`, `.csv`, `.json`) and Office documents (`.docx`, `.xlsx`, `.pptx`). Older Office formats are rejected; convert them first.
* Video and audio are not chat attachments. Add them through the Media pool, then reference them with `@`. See [Import media](/editor/import-media).

<Frame>
  <img src="https://mintcdn.com/goframe-fb816310/nCYzUacADHcqMX4R/images/editor/prompting-attachments.png?fit=max&auto=format&n=nCYzUacADHcqMX4R&q=85&s=9ac717fc469a9192c332732d681c3ecf" alt="Composer with a PDF and an image attached" width="800" height="420" data-path="images/editor/prompting-attachments.png" data-path="images/editor/prompting-attachments.png" />
</Frame>

## Use voice input

Click **Voice input** in the composer or press `⌘⇧M` (Ctrl on Windows) to dictate a prompt. Speech recognition runs locally in your browser; the first time you use it, a small model downloads ("Loading model… N%"). Pick the language from the button's menu (**Auto-detect**, **English** or **Russian**), speak, then click **Stop and transcribe**. The text lands in the composer for you to edit before sending.

## Set generation options

The **Generation settings** button (sliders) in the composer opens the options for image and video prompts:

* **Aspect ratio**: **Auto**, **16:9**, **9:16**, **1:1**, **4:3**, **3:4**, **4:5** or **21:9**.
* **Duration**: **Auto** or a clip length. The choices follow the selected video model (**4s**, **6s** or **8s** with the default model).
* **Audio**: **Auto**, **On** or **Off**.

Model-specific limits are listed in [AI models](/editor/ai-models).

<Frame>
  <img src="https://mintcdn.com/goframe-fb816310/nCYzUacADHcqMX4R/images/editor/prompting-generation-settings.png?fit=max&auto=format&n=nCYzUacADHcqMX4R&q=85&s=44e64eae7b95aeda030205459b6f2a47" alt="Generation settings popover with Aspect ratio, Duration and Audio chips" width="810" height="620" data-path="images/editor/prompting-generation-settings.png" data-path="images/editor/prompting-generation-settings.png" />
</Frame>

## Start from a scenario template

On the Home page, the template cards start a project with a ready-made scenario: **Talking head**, **Product promo**, **Explainer**, **Shorts cutdown**, **Photo slideshow** and **Captions & cleanup**. Attach starter files (a script, brief or media) on the same screen. See [Home and projects](/editor/home-and-projects).

## Continue

* [Meet GoFrame Editor Agent](/editor/agent)
* [Skills and workflows](/editor/skills)
* [AI models](/editor/ai-models)
* [Review proposals and approve actions](/editor/proposals-and-permissions)