caption_subtitle_generator / tests /test_subtitle_utils.py
DILEEPJAAT's picture
Upload 8 files
18e68a0 verified
Raw History Blame Contribute Delete
3.68 kB
import unittest
from subtitle_utils import (
Cue, TimedWord, build_cues, chunks_to_words, timestamp_chunk_coverage,
timed_words_to_text, to_json, to_srt, to_vtt, transcript_coverage,
)
class SubtitleUtilsTests(unittest.TestCase):
def test_word_chunks_generate_readable_srt_vtt(self):
chunks = [
{"text": " Hello", "timestamp": (0.1, 0.4)},
{"text": " there.", "timestamp": (0.45, 0.8)},
{"text": " How", "timestamp": (1.1, 1.3)},
{"text": " are", "timestamp": (1.3, 1.5)},
{"text": " you?", "timestamp": (1.5, 1.9)},
]
words = chunks_to_words(chunks, "Hello there. How are you?", 2.0, 42)
cues = build_cues(words, max_chars=42)
self.assertTrue(cues)
self.assertIn("Hello there.", to_srt(cues))
self.assertTrue(to_vtt(cues).startswith("WEBVTT"))
self.assertIn('"text"', to_json(cues))
def test_segment_chunk_is_split_for_caption_width(self):
chunks = [{"text": " This is a longer segment that should wrap into readable subtitle lines.", "timestamp": (0.0, 4.0)}]
words = chunks_to_words(chunks, "This is a longer segment that should wrap into readable subtitle lines.", 4.0, 28)
cues = build_cues(words, max_chars=28)
self.assertTrue(cues)
for cue in cues:
self.assertLessEqual(len(cue.text.splitlines()), 2)
def test_missing_chunk_timestamps_falls_back_to_full_text(self):
words = chunks_to_words([], "A small test transcript", 3.0, 42)
self.assertTrue(words)
cues = build_cues(words)
self.assertTrue(cues)
def test_very_long_token_is_not_dropped(self):
long_token = "x" * 150
chunks = [{"text": long_token, "timestamp": (0.0, 3.0)}]
words = chunks_to_words(chunks, long_token, 3.0, 42)
cues = build_cues(words, max_chars=42)
recovered = "".join(cue.text.replace("\n", "") for cue in cues)
self.assertEqual(recovered, long_token)
def test_srt_timestamp_format(self):
srt = to_srt([Cue(1.234, 62.5, "Timing")])
self.assertIn("00:00:01,234 --> 00:01:02,500", srt)
def test_missing_timestamp_chunks_are_not_dropped(self):
chunks = [
{"text": " Hello", "timestamp": (0.0, 0.4)},
{"text": " missing", "timestamp": (None, None)},
{"text": " timestamp", "timestamp": (None, None)},
{"text": " preserved.", "timestamp": (1.1, 1.6)},
]
words = chunks_to_words(chunks, "Hello missing timestamp preserved.", 2.0, 42)
rebuilt = timed_words_to_text(words)
self.assertEqual(transcript_coverage("Hello missing timestamp preserved.", rebuilt), 1.0)
self.assertEqual(timestamp_chunk_coverage(chunks), 0.5)
def test_partial_chunk_text_is_detectable_for_retry(self):
chunks = [
{"text": " This part is present", "timestamp": (0.0, 1.5)},
]
words = chunks_to_words(chunks, "This part is present but this sentence is missing", 4.0, 42)
rebuilt = timed_words_to_text(words)
self.assertLess(transcript_coverage("This part is present but this sentence is missing", rebuilt), 0.985)
def test_all_text_without_timestamps_is_preserved(self):
chunks = [
{"text": " One chunk", "timestamp": (None, None)},
{"text": " another chunk.", "timestamp": (None, None)},
]
words = chunks_to_words(chunks, "One chunk another chunk.", 3.0, 42)
self.assertEqual(transcript_coverage("One chunk another chunk.", timed_words_to_text(words)), 1.0)
if __name__ == "__main__":
unittest.main()