@vilyasik3

videoMD · tier A

M1dekz · 586 followers · 2026-07-12 09:22 · 12s

Audio track (35.4 KB), what ASR actually heard

16 kHz mono Opus. Ogg Opus does not play in Safari; Chrome and Firefox are fine.

eagle · fetched 2026-07-26 05:44 · 1.2 MB downloaded · 2 keyframes

plays

290

likes

comments

Topic

otherconf 0.00neutral

Fixed taxonomy enforced as a schema enum. Prose instructions did not hold, and label drift makes week-over-week comparison meaningless.

gpt-4.1-nano · prompt v4 · $0.000088 · 2026-07-26 08:41

Caption

Author-written, as posted.

No caption. True for 83.6% of the corpus.

Speech (ASR)

hallucinationru 74%cyrillic

faster-whisper large-v3-turbo behind Silero VAD, then a repetition filter.

Субтитры сделал DimaTorzok

rtx4090-24-02 · 26 chars · 5.4s speech of 12.7s

On-screen text (OCR)

easyocr

EasyOCR for Latin script, macOS Vision for Cyrillic-suspect frames.

BOTH

rtx4090-24-01 · 0 chars · 2 frames