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

# Galileo

> Evaluate video for visual glitches and prompt misalignment.

Galileo checks generated video for visual defects and whether it matches the
prompt used to create it. Use **`galileo-1.0`** as the model name in your evaluation.

## Capabilities

| Check            | What it detects                                                             |
| ---------------- | --------------------------------------------------------------------------- |
| Visual glitches  | Visual defects such as deformation, clipping, and inconsistent motion.      |
| Prompt alignment | Objects, actions, or details in the prompt that the video does not fulfill. |

With a prompt, Galileo runs both checks. Without one, it checks visual glitches only.

## Supported inputs

|         | Supported                                    |
| ------- | -------------------------------------------- |
| Video   | H.264 MP4, up to 15 seconds and 50 MiB       |
| Sources | Local file, public URL, or uploaded video ID |
| Prompt  | Optional, up to 2,500 characters             |

The SDK uploads local files automatically. See [video inputs](/guides/video-inputs)
for supported source formats and reusing an upload.

## Results

Each evaluation returns a summary and a list of findings. Visual-glitch findings
locate the affected frames and regions. Prompt-alignment findings identify the
relevant prompt text and rate the mismatch from 1 (minor) to 5 (requirement absent).

An empty list means no findings were reported by the completed checks. See
[evaluations](/guides/evaluations) for result fields and how to handle partial results.