bug: params.with_max_segment_length(1) doesn't produce word-level segments

[ProfHercules]

Describe the bug

Given this code:

model = w.Whisper.from_pretrained("tiny")
params = model.params.with_max_segment_length(1).build()

samples = [] # np.array of samples from pydub
model.context.full(params, samples)
for s in range(model.context.full_n_segments()):
    print(model.context.full_get_segment_text(s))

I would expect the output to be individual segments, but instead I just get normal sentences.
It doesn't seem like with_max_segment_length makes any difference.

I may be misunderstanding some things - my main goal is to just get word-level timestamps by outputting individual words.

To reproduce

No response

Expected behavior

No response

Environment

python: 3.11
platform: MacOS 13.3