> For the complete documentation index, see [llms.txt](https://docs.chatvideopro.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.chatvideopro.com/workflows/how-to-generate-b-roll-inside-premiere-pro-with-ai.md).

# How to Generate B-Roll Inside Premiere Pro with AI

Generate B-roll, establishing shots and cutaways inside Premiere Pro with AI. Pick a video model or a guided Studio workflow and save the clip to your project.

A missing establishing shot. A cutaway you never filmed. An atmosphere clip that would carry the transition. B-roll gaps are one of the most common blockers in an edit, and stock footage rarely matches the mood of your project.

Chat Video Pro lets you create that shot inside Premiere Pro: from a prompt, from a still, or from a frame of your own footage. The finished clip saves straight into your Premiere project. There is no browser tab and no export-and-reimport loop.

{% hint style="info" %}
**How AI is billed.** Video generation runs on your own fal.ai account and is billed by fal.ai at its price, only when you generate, with no markup from us. Before a paid video runs from the composer, a confirmation card shows the model and an estimated cost, and nothing is sent until you confirm. Check your spending in **Gear → Usage**. See [Pricing](https://docs.chatvideopro.com/getting-started/pricing.md).
{% endhint %}

<figure><img src="https://3726163538-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FOksMNVulq2ffwzQ32RfP%2Fuploads%2FW56fzAgGk3dKbdotgk8y%2Fgenerated-image-make-this-a-landscape-image-style-photorealistic-1.png?alt=media&amp;token=22e9a821-72dc-49e4-9b45-08e34401462c" alt="Seedance promotional illustration with AI video tools in Premiere Pro"><figcaption></figcaption></figure>

***

## Two ways to make B-roll

| You want                                                | Use                                                                                                                                                                                             | What you get                                         |
| ------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------- |
| Full control of the model, length, resolution and audio | **Generate Media** in the composer                                                                                                                                                              | A clip from your prompt, or from an image you attach |
| A polished shot without prompt engineering              | **Studio:** [Cinematic Lab](https://docs.chatvideopro.com/features/studio/cinematic-lab.md) → [Motion Director](https://docs.chatvideopro.com/features/studio/motion-director.md)               | A designed still, animated with a guided camera move |
| More angles of a shot you already have                  | **Studio:** [Multi-Cam](https://docs.chatvideopro.com/features/studio/multi-cam.md)                                                                                                             | New camera angles from an image or a short clip      |
| A bridge between two shots                              | **Studio:** [AI Transitions](https://docs.chatvideopro.com/features/studio/ai-transitions.md), or [Transition Mode](https://docs.chatvideopro.com/features/video-generation/transition-mode.md) | A moving shot from a start frame to an end frame     |

If you are new, start with the Studio route below. It does more of the prompt and model setup for you.

***

## Choose a video model

Pick the model by the hardest part of the shot. These are good starting points; the model picker always shows the current list, and [Supported Video Models](https://docs.chatvideopro.com/features/video-generation/supported-video-models.md) has the full specs.

| The shot needs...                                 | Start with                                                   | Why                                                                                                                                        |
| ------------------------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------ |
| A quick test, or sound and dialogue               | **Omni Flash**                                               | The default video model. Audio is generated with the picture on every clip. Draft at a low resolution, then step up for the take you keep. |
| Cinematic camera movement or a product move       | **Kling V3** (Kling 3.0) or **Kling O3**                     | Strong camera movement and polished motion                                                                                                 |
| A longer Kling clip, or several shots in one clip | **Kling 4.0**                                                | 3 to 30 seconds at 720p or 1080p, with Multi Shot                                                                                          |
| Natural motion with ambient sound, or native 4K   | **Seedance 2**                                               | Balanced all-rounder; its 4K resolution pill renders 4K without an upscale                                                                 |
| A clip longer than 20 seconds, with sound         | **Seedance 2.5**                                             | 4 to 30 seconds in one pass, up to 1080p                                                                                                   |
| A silent cinematic plate you will score yourself  | **Luma Ray 3.2**                                             | No audio on any tier                                                                                                                       |
| Action, sports or fast movement                   | **Hailuo 03**, or **Hailuo 03 Max** for near-realtime drafts | 2K with native audio; Max is the fast drafting tier                                                                                        |
| Ultra-wide 21:9 or 2:1 framing                    | **Flux 3**                                                   | Offers both 21:9 and 2:1                                                                                                                   |

***

## The workflow

### 1. Identify the gap

While cutting, note what is missing: an establishing shot, a cutaway, an atmosphere clip, a concept visual. Note the composition you need (wide, close-up, POV), the mood, whether there are people in it, and whether you want sound.

### 2. Capture a frame for context (optional)

Park the playhead on a nearby shot with the look you want to match and click [**Frame Capture**](https://docs.chatvideopro.com/getting-started/interface-overview/frame-capture-button.md). The frame attaches to the composer.

* With one image attached, most models run **image-to-video** and use it as the start frame, so the new shot continues your footage's look.
* To use the frame as a style or subject reference instead, pick a **Reference** model. See [Reference Mode](https://docs.chatvideopro.com/features/video-generation/reference-mode.md).

### 3. Turn on Generate Media and set up the shot

Click the [**Generate Media**](https://docs.chatvideopro.com/getting-started/interface-overview/generate-media-button.md) toggle in the composer, choose a video model, and set the aspect ratio, duration and resolution. Models with an audio pill let you switch sound on or off; Omni Flash always generates audio.

### 4. Write the prompt

A strong B-roll prompt has four parts:

```
[Subject/scene] + [Camera movement] + [Lighting/mood] + [Style]
```

Example:

```
Aerial drift over a coastal town at golden hour. Slow, gentle movement. Warm evening light. Cinematic, realistic documentary style.
```

Describe what you see, not what you want it to mean. "Rainy street at night with neon reflections" works better than "moody urban atmosphere." Click the **AI Optimize** spark next to the Generate Media toggle to have your draft rewritten for the model you picked; you review the rewrite before it replaces your text.

### 5. Generate

Send the prompt and confirm the card. Generate two or three takes with small prompt changes. AI generation varies, so give yourself options to choose from.

### 6. Bring it into Premiere

Click the download button on the finished clip. Inside Premiere, Chat Video Pro saves the file and imports it into your project, ready to drag onto the timeline. Every result also stays in your [Library](https://docs.chatvideopro.com/getting-started/interface-overview/library.md) for later.

***

## The Studio route: design the frame, then animate it

Use this when the look matters more than speed.

1. Open **Studio → Cinematic Lab**. Describe the scene, choose the camera, lens, focal length and aperture, pick the aspect ratio of your deliverable, and generate up to 4 options.
2. Click **Done** on the strongest frame. It goes into your chat.
3. Open **Motion Director**, pick that still from **Recents**, choose a camera move (push in, dolly, orbit, crane, handheld, drone or reveal) and click **Generate**.
4. Download the result to save it into your Premiere project.

Need coverage around the same moment? Run the still or a short clip through **Multi-Cam** for new angles. Need to join two shots? Use **AI Transitions**. See [Studio](https://docs.chatvideopro.com/features/studio.md) for every workflow, or the [Studio page](https://www.chatvideopro.com/studio/) for an overview.

***

## Beyond basic B-roll

**Reference Mode: the same subject across several clips.** Reference models keep a product, character or location consistent across generations. For example, Kling O3 Reference takes 1 to 7 reference images, and Kling 4.0 Reference takes up to 10 images, 7 elements and 5 video clips. See [Reference Mode](https://docs.chatvideopro.com/features/video-generation/reference-mode.md).

**Transition Mode: the moment between two frames.** Attach two frames from your timeline and the model treats them as a start frame and an end frame, then generates the motion that connects them. See [Transition Mode](https://docs.chatvideopro.com/features/video-generation/transition-mode.md), or use Studio AI Transitions for guided transition styles.

**Generative Extend: a clip that ends too soon.** Attach the clip and add 2 to 15 seconds of new footage with sound. See [Generative Extend](https://docs.chatvideopro.com/features/video-generation/generative-extend.md).

**Multi-Cam: a new angle on a shot you have.** Generate a close-up, a reverse angle or a different focal length from an image or short clip, without a second camera. See [Multi-Cam](https://docs.chatvideopro.com/features/studio/multi-cam.md).

***

## Tips for best results

**Match your footage, not stock.**\
Start from a captured frame so the new shot carries your edit's color, lighting and composition.

**Draft fast, finish high.**\
Find the prompt on a faster or lower-resolution setting, such as Omni Flash at a low resolution or Hailuo 03 Max, then regenerate the winning prompt on a higher-quality model or setting.

**Decide on sound early.**\
For atmosphere shots, a model with ambient audio can give you a usable bed. For shots you will score yourself, a silent model like Luma Ray 3.2 keeps the track clean.

**Describe camera movement specifically.**\
"Slow push-in", "FPV drone shot" and "static wide" produce very different results. Both Kling and Seedance respond well to cinematic camera language.

**Upscale last.**\
Approve the creative version first, then run [Video Upscaling](https://docs.chatvideopro.com/features/studio/video-upscaling.md) on the final take.

***

## Next steps

* [**Supported Video Models**](https://docs.chatvideopro.com/features/video-generation/supported-video-models.md) — Full specs, duration limits, and audio capabilities for every model
* [**Reference Mode**](https://docs.chatvideopro.com/features/video-generation/reference-mode.md) — Keep subjects and characters consistent across multiple generations
* [**Transition Mode**](https://docs.chatvideopro.com/features/video-generation/transition-mode.md) — Generate cinematic motion between two keyframes
* [**Frame Capture**](https://docs.chatvideopro.com/getting-started/interface-overview/frame-capture-button.md) — Capture frames from your timeline as generation context
* [**First 5 Things to Try**](https://docs.chatvideopro.com/getting-started/first-5-things-to-try.md) — The quickest wins in Chat Video Pro

***

**Related Workflows:**

* [AI VFX Inside Premiere Pro](https://docs.chatvideopro.com/workflows/how-to-create-ai-powered-visual-effects-in-premiere-pro.md)
* [Story Editing with AI in Premiere Pro](https://docs.chatvideopro.com/workflows/how-to-cut-videos-faster-with-ai-assisted-story-editing-in-premiere-pro.md)
* [AI Thumbnails in Premiere Pro](https://docs.chatvideopro.com/workflows/how-to-generate-high-ctr-thumbnails-inside-premiere-pro-with-ai.md)
