>>/29366/
If you really want to go that route, we need to start with the vowel sounds, long and short

a - ah ayy
e - eh eeh
i - ih eye
o - oh owh
u - uh ooh

That's already ten phonetic characters

b - buh
c - ceh
d - duh
f - fff
g - guh
h - heh
j - juh
/k - kuh ceh
l - lll
m - mmm
n - nnn
p - puh
/q - ceh ooh uh
r - rrr
s - sss
t - teh
v - vvv
/w - wuh - ooh uh
/x - eh ceh sss
/y - yuh - eeh uh
/z - teh sss

Then we can eliminate characters that can be spelt phonetically with other characters that's another fifteen

ch th ng 

Plus three special characters.

a - ah
? ai - ayy - ah eeh
e - eh
? ee i - eeh
i - ih
? i - eye - uh eeh
o - oh
? ow - owh - oh ooh
u - uh
? oo - ooh

b - buh
c - ceh
d - duh
f - fff
g - guh
h - heh
j - juh
l - lll
m - mmm
n - nnn
p - puh
r - rrr
s - sss
t - teh
v - vvv

? - ch
? - th
? - ng

Last we assign our characters to the phonetic sounds, then we have our cypher. If we learn the alphabet we should be able to read and write the language which will sound like English, although slightly broken like a foreigner might use. However in theory we/I should be able to write a translation script in python/whatever to aid learners.
As for what characters you would like assigned, we could stick to what's readily available on the keyboard or even toss in a few special characters, just to fuck with things more. The standard qwerty already has many characters on it like `¬%^&*_+{}~@:<>/#[]