Image Input
Halo supports sending images directly in conversations. The AI can recognize and analyze image content and respond accordingly.
How to Add Images
Any of three methods work:
- Click the "+" button on the left side of the input box and select a file
- Drag and drop an image file directly into the input box
- Use Ctrl/Cmd + V to paste a screenshot from your clipboard
Supported Formats and Limits
| Item | Details |
|---|---|
| Supported formats | PNG, JPEG, WebP, GIF |
| Maximum file size per image | 20 MB |
| Maximum images per message | 10 |
Important: A Multimodal Model Is Required
Non-multimodal models will throw an error
Image input only works with multimodal models, such as Claude 4.5, Claude 3.7 Sonnet, and GPT-4o.
When using a model that does not support vision (such as some domestic model APIs), uploading an image will cause an error. Because the context already contains image parameters, you must start a new conversation to continue using Halo normally.
To switch to a multimodal model: use the model selection menu in the top-right corner.
Typical Use Cases
- Screenshot the UI and ask the AI to analyze the design or identify code errors
- Upload a chart and ask the AI to interpret the data
- Send a design mockup and ask the AI to generate the corresponding code
- Photograph a document and ask the AI to extract or translate the text