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

# Compare AI models side by side in the Playground

> Run any image, video or audio model in the Percify Playground, or give two models one prompt in Compare. Each run shows its credit price first.

The Percify Playground lets you run image, video and audio models one at a time or two at a time. Pick a model, fill in its inputs and press **Run**, or open **Compare** to send the same prompt to two models and watch both results side by side. Every run shows its price in credits before you start, and a run that fails is refunded.

Open it at [app.percify.io/home/playground](https://app.percify.io/home/playground). You can browse the models without an account. You need to sign in to run one.

## How do I run a single model?

<Steps>
  <Step title="Find a model">
    Open **Playground** in the sidebar. Sort the catalog with **Discover**, **Trending** or **Latest**, type in **Search models**, or use the video, image and audio icons to filter by what the model makes. See [AI models in Percify](/models) for the full list.
  </Step>

  <Step title="Fill in the inputs">
    On the model's page, the **Playground** tab shows an **Input** form built from that model's own settings, such as the prompt, a reference image, length or resolution. You can also edit the same values as **JSON**. **Reset** puts every field back to its default.
  </Step>

  <Step title="Check the price and run">
    The header shows the price for your current settings, and the button reads **Run** with that number. If a required field is empty, Percify tells you which one before anything is charged.
  </Step>

  <Step title="Save the result">
    The result appears under **Output**. Use **Download** to save the file or **Copy link** to copy its URL.
  </Step>
</Steps>

Each model page also has an **Examples** tab with sample outputs (press **Use input** to load an example's settings into the form) and a **README** tab that lists every input and explains how the model is priced. Some model pages also have an **API** tab with ready-to-copy requests. To run models from your own code, see the [API reference](/api-reference/introduction); API access is included on the Scale and Ultra plans.

## How do I compare two models side by side?

<Steps>
  <Step title="Open Compare">
    In the Playground, press **Compare** at the right of the header.
  </Step>

  <Step title="Choose what to test">
    Under **What do you want to test?**, pick **Video models** or **Image models**. Under **Comparison type**, pick **Two models** to test two models on one prompt, or **Director mode** (explained below).
  </Step>

  <Step title="Pick the models">
    Choose **Model A** and **Model B**. If either model needs a starting image, upload one under **Reference image**.
  </Step>

  <Step title="Write the prompt and set the shape">
    Type your **Prompt**, then set **Ratio** (9:16 or 16:9), **Quality** (**Cheapest**, **Balanced** or **Best**) and, for video, **Duration**. Compare only offers durations that both models can make exactly, so the two clips match in length.
  </Step>

  <Step title="Run both">
    The total reads "credits for both". Press **Compare**. Each side shows its own progress, time taken and credit cost.
  </Step>
</Steps>

Compare can only give a model a prompt and one reference image. Models that need an audio track, a source video or two different images are left out of its model lists.

### What is Director mode?

Director mode keeps the model the same on both sides and compares two ways of writing the prompt. You pick one model, then a **Director A** and a **Director B**. Each director is an AI language model that rewrites your idea before it goes to the video or image model, in one of two styles:

| Style               | What the director writes                                                                      |
| ------------------- | --------------------------------------------------------------------------------------------- |
| **Enhanced prompt** | Your idea rewritten as a detailed, production-ready prompt                                    |
| **JSON direction**  | A structured treatment covering scene, camera, lighting and pacing, sent to the model as JSON |

The directions each side used are shown under the results, so you can see what produced each clip.

### What can I do with a finished comparison?

* **History** keeps every comparison you run, so you can open a past one later or remove it.
* The share button creates a public link to the comparison. Press it again to revoke the link, which stops the old link working.
* When both sides are done, **Create 9:16 cut** or **Create 16:9 cut** renders a branded side-by-side video at 1080×1920 or 1920×1080. You can edit the label over each clip and choose whose sound it keeps: **Left clip**, **Right clip** or **Both**. Then use **Download** to save it.

## How much does the Playground cost?

The Playground spends the same credits as the rest of Percify, on every self-serve plan. The price depends on the model and on the settings that move its price.

| What you run           | What you pay                                                                                                                                                                                     |
| ---------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| One model              | The price on the **Run** button. Some models charge a flat price per run. Others charge by clip length, audio length, resolution or quality, and the price updates as you change those settings. |
| Compare, two models    | The price of each model's run, added together. Each side is charged when it starts.                                                                                                              |
| Compare, Director mode | The two model runs plus 1 to 3 credits for each director.                                                                                                                                        |
| Marketing cut          | No credits                                                                                                                                                                                       |

Credits come back automatically when:

* a run fails, or a stuck run is stopped
* one side of a comparison fails (you pay only for the side that ran)
* a director fails to write its direction

Prompts and images are checked by content moderation before anything is charged. A rejected input shows "Prompt rejected by content moderation" or "Image rejected by content moderation" and costs nothing.

If your balance is too low, Percify shows how many credits the run needs and how many you have. See [How Percify credits work](/percify/credits) and [Not enough credits](/troubleshooting/not-enough-credits).

<Tip>
  In Compare, if your settings cost more than the cheapest option, a **reset to cheapest** link appears next to the total. Use it to test prompts before paying for higher quality or longer clips.
</Tip>

## Related

<CardGroup cols={2}>
  <Card title="AI models in Percify" href="/models">
    Every image, video and audio model in the catalog, with credit prices
  </Card>

  <Card title="How to write prompts" href="/guides/prompt-engineering">
    Prompt advice that fits how Percify's models take input
  </Card>

  <Card title="How Percify credits work" href="/percify/credits">
    Plans, monthly credits and what a run costs
  </Card>

  <Card title="Percify API" href="/api-reference/introduction">
    Run the same models from your own code
  </Card>
</CardGroup>

Browse the public catalog at [app.percify.io/playground](https://app.percify.io/playground).