Thank you. I hadn't heard of this concept before. I think it could be relevant here.

If the encryptor offers multiple inputs, one a dummy "sure, here is my password, officer" HTML and passphrase, the other the sensitive one this scheme could work. Hiding both in a single blob, plus some noise, could work in the browser too. Seems it might offer some degree of plausible deniability. Some investigation of the decrypted dummy HTML might reveal it isn't large enough to account for the size of the blob. This doesn't seem like an insurmountable obstacle though. If the tool hides an arbitrary number of inputs in blobs that are indistinguishable you'd seem to have a great deal of deniability.