Demo response different from runtime response - google-cloud-platform

Question

I'm trying to use the Google Cloud Vision API to OCR this image:
I'm using the following code the make the request:
const resp = await fetch(
`https://vision.googleapis.com/v1/images:annotate?key=${KEY}`, {
method: 'POST',
body: JSON.stringify({
requests: [{
image: {content: encoded},
features: [{type: "TEXT_DETECTION"}],
}]
}),
});
This works but there is some information missing from the result. If we look at the text field:
Dog Search
D G O OD D ODG O O D D O
O D O O G G G D O D G OG G
OGD GOGD GO G GO G D
D D D G D DO DOO G D O O
O DGOGG D O O G G O O D
DOG
Here's that visualized:
There are boxes around the characters which were recognized. But, if we put this image into the gcv demo application, we get this instead:
And this is what text looks like:
Dog Search
D GOOD D 0 D GOOD DO
0 D 0 0 G G G DOD GO GG
o G O G D 0 0 D G 0 0 D D D
D G D o o o G G o o G D Go
0 G D G O G D G O G G O G D
D D D G D DO DO O G D 0 0
O D GO G G D 0 0 G G 0 0 D
DOG
Here's a gist with the requests + responses. I'm authenticating using a API token.
Why are the responses different? The requests are slightly different but not in a way which should affect the output. Right?