6 Commits

Author SHA1 Message Date
763999f3a9 feat: anti-hallucination params + retranscribe script for fixing broken transcripts
- transcribe.py: add --max-context 0, --entropy-thold 2.4, --max-len 60,
  --suppress-nst, --no-fallback to whisper.cpp to prevent hallucination loops
- transcribe.py: remove interactive quality gate (runs unattended now)
- run.bat: remove pause prompts for unattended operation
- retranscribe_tail.py: new script that detects hallucination bursts in SRT
  files, extracts and re-transcribes only the affected audio segments, then
  splices the result back together. Drops segments that re-hallucinate
  (silence/music). Backs up originals to transcripts/backup/.
- fix_hallucinations.bat: Windows wrapper for retranscribe_tail.py

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-24 21:17:14 +02:00
45e72bc89b feat: adauga --modules filter si la download.py
Parametrul din run.bat (ex: 4-5) era transmis doar la transcribe.py.
Acum download.py primeste acelasi filtru si descarca doar modulele
specificate. Sintaxa acceptata: '4-5', '1,3', '1-3,5'.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-24 02:10:33 +02:00
7b18e8fc41 fix(run.bat): evita Microsoft Store Python stub care termina cmd.exe
Store stub-ul pentru python.exe termina procesul batch cand e apelat
direct. Fix: foloseste 'py' launcher (safe) sau detecteaza python.exe
real via 'where | findstr /v WindowsApps', fara a executa python
in contextul check-ului. Toate apelurile python -> !PYTHON_CMD!.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-24 02:06:42 +02:00
e83bd74813 feat: switch to CPU-only whisper build (no GPU on this machine)
- setup_whisper.py: descarcă build CPU din release-urile oficiale,
  sare peste Vulkan/CUDA/OpenBLAS
- run.bat: elimină env var GGML_VK_PREFER_HOST_MEMORY și check-ul Vulkan SDK
- transcribe.py: --no-gpu era deja setat

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-24 02:01:39 +02:00
696c04c41c chore: normalize line endings from CRLF to LF across all files
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-24 01:53:35 +02:00
bbc5884545 NLP Master: pipeline download + transcribe + summarize
- run.bat: one-click pipeline (download, convert, transcribe)
- download.py: fetch audio from course platform
- transcribe.py: whisper.cpp batch transcription (CPU, WAV 16kHz)
  - MP3->WAV conversion via ffmpeg
  - --modules filter for splitting work across machines
- summarize.py: generate summaries from transcripts
- setup_whisper.py: auto-download whisper.cpp, ffmpeg, and model
- Medium model (q5_0) instead of large to avoid VRAM crashes

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-24 01:37:13 +02:00