Prompting with Image Not Working as Expected

[recursivecodes]

Trying to prompt with an image, per the README - and no matter what type of image I send, I get a response that indicates that the model sees 'object Object' instead of the actual image data.

Ex:

The image "[object Object]" is a common placeholder in web development, particularly in JavaScript. It's a placeholder that represents an object variable that hasn't been assigned a specific value yet. 

....etc

I've tried HTMLVideoElement, VideoFrame and Blob, all produce similar results.

Version 134.0.6998.89 (Official Build) (arm64)

[recursivecodes]

My session is created as follows:

const options = {
  expectedInputs: [
    { type: "image" }
  ]
};
this.aiSession = await ai.languageModel.create({
  ...options,
  monitor(m) {
    m.addEventListener("downloadprogress", (e) => {
      console.log(`Downloaded ${e.loaded} of ${e.total} bytes.`);
      self.aiModelDownloadProgress = Math.round((e.loaded / e.total) * 10000) / 100;
      if (e.loaded === e.total) self.aiModelDownloadProgress = null;
    });
  }
});

[tomayac]

Hi Todd,

The explainer is ahead of what's actually implemented in Chrome. We'll announce it on the early preview program once these APIs can be tested.

Cheers,
Tom