The reference, with one thing made explicit: the vowels i and u disappear between voiceless consonants and at the end of a word after s. 'Desu' is 'des', 'shita' is 'shta'. It is the feature that makes Tokyo Japanese sound clipped, and the one Kansai does not do.
Where: Tokyo and the standard
The way in: Whisper every u after s: 'desu', 'masu' become des, mas.
Try a word
Anything the spelling engine can transcribe, the accent can move. Slow it down to hear the shape.
Pitch accent, not stress: each word has one place where the pitch drops, and that place is the word's identity. 橋 hashi falls after ha, 箸 hashi falls after shi. Sentences are level with a fall at the end; the question particle ka rises.
What moves, one rule at a time
In the order the mouth applies them. Every before-and-after is playable — click the IPA. The pattern is the rule as the engine runs it, for the curious.
Vowels
u drops after s at the end of a word
([sɕ])ɯ(?=$| ) → $1
When: after s or sh, at the end of a word
How: Say the s and stop. Do not voice anything after it.
Vowel devoicing. The mouth still shapes the u, but the vocal cords do not turn on, so nothing is heard. The spelling keeps it; the ear does not.
Word
Reference
Accent
desu
masu
arimasu
i and u go silent between voiceless consonants
([kstɕɸçhp])[iɯ](?=[kstɕɸçhp]) → $1
When: between two of k, s, t, sh, f, h, p
How: Whisper the vowel or skip it entirely — either is native.
The same devoicing inside a word. 'Shita' has two syllables on paper and sounds like one and a half.