> Index of Phersa pages for AI agents: https://phersa.com/llms.txt

# Create

Create builds your video one clip at a time. Each clip starts as an image of your persona, becomes a video, and can take your persona's voice. Every Generate button shows the provider's price before you run it.

## Start a project

1. Open **Create** and start a **New Project**.

2. Give it a name and pick a persona. Every project needs one.

3. The first clip starts with your persona's photo as its reference image.

Your AI agent can do this part for you over [MCP](https://phersa.com/docs/mcp), writing every clip from a proven video. You can also load clips from a JSON file with **Import**.

## The three columns

- **Transcript:** your clips in order, each with its dialogue. **Add clip** adds another.
- **Preview:** the selected clip's active image at full resolution; video takes open in a player.
- **Inspector:** the settings for the selected clip, in three tabs: **Image**, **Video** and **Voice Swap**.

## Generate a clip

1. ### Image

    Write the image prompt, pick a model, aspect ratio and resolution, and choose **Generate image**.

2. ### Video

    Write the video prompt and the dialogue, pick a model, duration, aspect ratio and resolution, and choose **Generate video**. Videos always include sound.

3. ### Voice Swap

    Pick a voice, keep or remove the background sound, and choose **Swap voice**. It needs a video with dialogue and an ElevenLabs key.

Generations run in the background, so you can keep working. Each one becomes a take on the clip. Click a take to preview it; you choose which takes to use when you compile.

## Models

| Type | Models |
| --- | --- |
| Image | Nano Banana 2 (default) · GPT Image 2.5 Sunburst · Grok Imagine Image 2.0 |
| Video | Gemini Omni 1.1 Flash (default) · Grok Imagine Video 1.5 |
| Voice Swap | ElevenLabs Multilingual v2 (recommended) · English v2 |

Change your defaults with the gear icon in the project header (**Default Create Settings**). Which models you can use depends on the providers you connect; see [BYOK](https://phersa.com/docs/byok) and the [model prices](https://phersa.com/byok#models).

## What a generation costs

Each Generate button carries the provider's published price for that clip's settings in its label, such as **Generate video** followed by the price. The provider bills your own account; Phersa adds no markup. When a provider publishes no price for a model, the button says **Generate image** with no amount. Krea does not publish a price for GPT Image 2.5 Sunburst. On fal and Atlas Cloud, GPT Image 2.5 Sunburst is billed by tokens, so the button shows the output image's price for the chosen aspect ratio and quality followed by **\+ inputs**: the prompt and each reference image add a little more. With the aspect ratio set to Auto the model picks the size, so no amount is shown.

Generating spends money on your provider account. Check the model, resolution and length before you choose Generate.

## Compile to Edit

When your clips have the results you want, choose **Compile**, pick the results to use and name the video. They open in the **Media** panel of a new [Edit](https://phersa.com/docs/edit) project, ready to place on the timeline, cut, caption and export.

Source: https://phersa.com/docs/create