# AI assistant: edit with words

> How to edit a reel with plain-word commands — “make it more dynamic”, “cut to 30 seconds” — what the assistant can and can't do, and what it costs.

Source: https://docs.slicer.media/en/guide/ai/assistant/

The AI assistant edits the montage from a command written in plain words. It reads the reel as text — clips, speech, cut words, graphics — and picks the right edits: cut repetitions, reorder clips, fix captions, add emoji, a plaque or a hook. The edits are applied by the same tools as manual editing, so the result shows in the preview right away and can be undone in one step.

**Demo (Web):**

1. Press “AI assistant” in the main set — the “AI” mode opens to the right of the preview
2. Click the suggestion “Add emoji to key words”
3. While the assistant thinks, a spinner turns on the send button
4. Done: the list shows the command and what changed, emoji appear in the preview
5. Click “Remove filler words (free)” — “Filler words are cut.” or “No filler words found.”

## Giving a command

### 1. Open the “AI” mode

Press **AI assistant** in the main toolbar. On a wide screen the panel opens to the right of the preview, on a phone — below it.

### 2. Write a command

Describe what to change — up to 500 characters. `Enter` sends the command, `⇧ Enter` (Mac) / `Shift Enter` (Windows) adds a line break. Or click a ready suggestion: **Make it more dynamic**, **Remove repetitions and weak phrases**, **Cut to 30 seconds**, **Add emoji to key words**, **Write a catchy hook**, **Fix punctuation in captions** — it is sent right away.

### 3. See the result

Under the field you'll see the command and the assistant's short answer — what it changed. The preview already shows the new edit. Don't like it — **Undo** (`⌘ Z` (Mac) / `Ctrl Z` (Windows)): the whole command is undone in one step.

## What the assistant can do

| Task | Example command |
| --- | --- |
| Cut repetitions, slips, weak phrases | “Remove repetitions and keep the strongest ideas” |
| Shorten the reel to a length | “Cut to 40 seconds, keep the ending” |
| Bring back what isn't in the reel | “Bring back the phrase about the first results” |
| Remove or reorder clips | “Start the reel with the outcome and move the intro after it” |
| Fix caption text | “Spell the brand name correctly: Slicer” |
| Highlight key words with accents | “Highlight the numbers and names as accents” |
| Add or remove emoji | “Add emoji to key words” |
| Add a plaque: title, number, list, quote, name | “Add a plaque with the speaker's name and role at the start” |
| Add a camera zoom on an important phrase | “Add a zoom on the main idea” |
| Write or remove a hook | “Write a catchy hook about editing speed” |
| Add a media card with an illustration | “Show a card with a cup of coffee when I talk about the morning” |
| Move captions to the top, center or bottom | “Move the captions to the top” |
| Cut pauses by silence | “Remove pauses longer than a second” |
| Tune a clip: speed, volume, effect, transition, filter, animations | “Speed up the footage in the middle twice” |
| One transition, effect or filter for the whole reel | “Put a fade on every cut” |
| Carry one clip's look over to others | “Make all clips look like the first one in color” |
| Add or change a [title](https://docs.slicer.media/en/guide/editor/titles/) | “Add a title ‘Three simple steps’ at the start” |
| Place [markers](https://docs.slicer.media/en/guide/editor/markers/) with notes | “Put a marker at every change of topic” |
| Put a source into the reel as a new clip | “Add the second video to the end of the reel” |
| Change the cover or the aspect ratio | “Use the frame at the fifth second for the cover” |

The assistant works with the words actually spoken in the reel: it can cut, bring back or reorder them, but it doesn't add new speech. When shortening a reel, it removes the weakest parts and tries not to cut a thought in the middle.

## What the assistant doesn't do

Music, your own photos and videos over the frame (it can only move ones already placed), keyframes, Adjust, freeze frames, audio and video processing, voice recordings and the caption style are not changed by the assistant — they have their own tools on the timeline and in **Settings**. A clip's mask, background and framing it can only copy from another clip. Caption translation is a separate [Translate](https://docs.slicer.media/en/guide/editor/translate/) feature. If a command can't be done, the assistant explains it in the answer and changes nothing — no tokens are charged.

## Price

A command is charged **at cost** — exactly what the AI service charges for this answer, with no markup. The exact usage isn't known in advance: it depends on the length of the reel and on the command itself. So above the input field the price reads like “≈ X tokens, at most Y” — what a command usually costs and the limit it never goes over.

-   While the command runs, the price is reserved on your balance, rounded up to a whole token — so you need at least one free token.
-   The actual price is charged, and **only if something changed**. If there are no edits, the answer says **No changes — no tokens charged.** If the assistant didn't respond, the reserve returns to your balance.
-   Fractions of a token add up on the project and are charged once they make a whole token.

How at-cost pricing works — [Tokens, packs and the free reel](https://docs.slicer.media/en/guide/pricing/#at-cost).

**Remove filler words (free)** works without AI: it cuts one-word fillers like “um” and “uh” across all clips. If there are none, you see **No filler words found.**

> **Note.** Undoing a command brings the reel back, but not the tokens: the command was already done. Every command is recorded in the [project history](https://docs.slicer.media/en/guide/projects/history/).

## Writing commands

-   Say **what** you want to get: “Make the reel more dynamic, no pauses” works better than “cut word number five”.
-   Give the length in numbers: “Cut to 30 seconds”.
-   Point to the place with words from the speech: “Add a zoom when I talk about the result”.
-   Split a big change into several commands — each is easier to check and undo.
-   Describe a card concretely: “a cup of coffee on a table” — the AI draws the image on export.

## Good to know

-   The assistant sees the current reel with all your edits, even unsaved ones.
-   While a montage or export is running, commands aren't accepted — wait until it finishes.
-   If the AI is unavailable, you see **The AI assistant is unavailable** — try again later.
-   The assistant picks emoji from the set the renderer supports — so not every emoji from your request ends up in the reel.
-   The assistant's edits are regular edits: you can continue by hand on the timeline and in the **Elements** mode.

## What's next

-   [Elements: hook, plaques, cards, emoji, zooms](https://docs.slicer.media/en/guide/editor/elements/)
-   [Captions](https://docs.slicer.media/en/guide/editor/captions/)
-   [Operation prices](https://docs.slicer.media/en/guide/help/prices/)