SwiftKey should not predict the same word after Flowing if you already deleted that word
Trace out what you think is "clown" on the keyboard. SwiftKey inserts "clean" instead. Erase, trace it again. SwiftKey inserts "clean" again. Erase, curse out the phone, trace it again, this time carefully. SWIFTKEY INSERTS THE SAME *!@#(*!@$ WRONG WORD AGAIN! UGH! Manually type it out, letter by letter like some hairy primate living in 1999.
If the suggestion presented was deleted already, FOR THE LOVE OF GOD, DO NOT SUGGEST IT AGAIN for this session! I clearly didn't want it the first time (or second or THIRD TIME.) Please refer to the definition of "insanity" for more information.
Yes, sometimes the intended word is available to click, but most times it's not there at all. Instead there are three or four other words that are not even close to the intended word. Fat fingers, or whatever.
Thanks. Great app otherwise.
Hi Lisa,
I'm wondering how you're getting this new experience. We changed the way Flow works . When you Flow a word, you should see three options.
I tried to Flow "clown" and I got clean - but the left prediction bar had "clown"
Have you not updated your SwiftKey app recently?
This happens on the newest version, the new flow is actually the problem.
I tried writing "Test", and it was changed for "Tatt" every time. 2/6 times "test" was one of the suggestions, but was always changed for tatt automatically after flowing.
Not sure why "Tatt" is even suggested as I clearly swiped over the E, and was not touching A.
As the swiftkey-team doesn't care about the forum-post (as moderators will tell you) you have to create a feature-request to make them aware that the keyboard is full of bugs (or as they like to call it "new features").
You can create a feature-request here:
https://support.swiftkey.com/hc/en-us/requests/new
The new flow experience doesn't change the way the predictions work. I think the real problem here is that the predictions are bad at suggesting words that you meant to flow because errors during flowing are different from typing, which is what the predictions seem to be rooted in. Flowing the area from p->r is a serious problem and I often end up typing instead. If I accurately flow "it" my predictions are "if | it | it is" but the closest words to what I actually flowed would be "out | it | or | our".
A trick I've come up with is I will flow words wrong to get them right. "in" I flow u->n and "on" is p->n. Another trick is to artificially create stepping points by arching between letters instead of going directly from letter to letter. Like "pour", if I go p->r I get "poor", but if I arch po->u->r I get pour.