How can I pass binary image data from an Open AI image generator subagent to Telegram in n8n?

Below is one solution approach:

Use n8n’s Native Binary Images Option

n8n lets you control how binary images are passed between nodes. Instead of having the agent inject the raw binary into JSON (which breaks the JSON), you can configure the agent so that image files are automatically passed as image‐type messages. In practice, this means enabling the binary images option on the agent. With this setting turned on, the agent won’t try to embed raw binary into the JSON payload, and you’ll receive a proper binary property that you can forward to your Telegram node.

Steps to Implement

1. Enable Binary Images in the Agent:
   In the agent’s configuration (or via environment variables if self-hosting), enable the option that automatically passes binary images as image type messages. This prevents the image binary from getting mangled into the JSON structure.

2. Ensure the Subworkflow Returns Binary Correctly:
   Make sure your imageCreator workflow returns the file as binary data. Use nodes like the [Convert to File] or [Extract From File] if needed so that the output is a proper binary property rather than a plain JSON string.

3. Pass the Binary to Telegram:
   When calling the Telegram node, refer to the binary property (for example, something like:

   {{$json["binaryPropertyName"].data}}
   

   ) instead of trying to parse it as JSON. This way, Telegram receives the binary file as expected.

Summary

By turning on the binary images option in your agent you ensure that image files are passed separately (as binary data) instead of being injected into the JSON payload. This lets you easily pipe the output from your imageCreator workflow directly into your Telegram node, avoiding JSON parsing issues.

This solution aligns with n8n’s built-in binary handling functionality and prevents breaking the JSON when dealing with files.