좌우 분할: 왼쪽=음성(Whisper 정밀+한국어), 오른쪽=화면(Gemini 한국어). 각각 SRT 다운로드.
음성 한국어는 기존 전사+LibreTranslate 재사용, 화면은 Gemini 신규(transient).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
'Gemini 자막' 버튼으로 유튜브 URL을 Gemini에 보내 화면/음성 자막을 추출+한국어 번역,
[00:00] 한국어 세그먼트로 저장 → SRT 내보내기가 곧 한국어 SRT. 제거된 OCR의 대체.
무료 티어(하루 8시간) 사용, 타임스탬프는 ~1초 근사.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Tesseract 기반 OCR이 실제 쇼츠의 스타일 자막을 거의 못 읽어 실효성이 없고,
URL→Gemini 방향으로 전환하기로 해 h-lab의 OCR 연동을 제거한다.
- 컨트롤러 POST /{id}/ocr 제거
- CurationService.ocrScreenSubtitles 제거
- ChannelService.ocrFromCached/joinSegmentText 제거(persistScript는 전사가 계속 사용)
- rework.html '화면 자막 OCR' 버튼·crop/fps 입력·ocrScreen() 제거
- docs/python-service/ocr_video_endpoint.py(videocr 예시) 삭제
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
영상 박힌 자막 → SRT/세그먼트(POST /ocr_video). 응답을 /transcribe 와 동일 형식
({language, segments[{start,end,text}], srt})으로 맞춰 h-lab 연동 재사용 가능하게 함.
h-python 서비스에 배포할 참고 코드.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
재가공 화면 '원본 다운로드' 버튼으로 yt-dlp 로컬 실행 → 서버 캐시 →
기존 Python 전사/렌더 파이프라인 재사용. 전사·렌더 양쪽에서 수동 업로드 제거.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Light-default + dark toggle, refined SaaS components, phased rollout across
all SSR pages. Approved dashboard mockup as the visual reference.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Phase 3: remove "no-talk" gaps (Whisper-segment based, not audio silencedetect
which finds nothing under background music) and render a trimmed (+speed) video
via ffmpeg, with subtitles remapped to match.
- KeepIntervalPlanner + TimelineRemapper (pure, unit-tested): keep/remove plan
from segments (pad/minGap) and timestamp remap f(t)=t-removedBefore(t)
- GET /{id}/trim-plan (preview: keep/remove/remapped segments/kept duration)
- POST /{id}/render (multipart: file,pad,minGap,speed) -> proxy Python /render
(ffmpeg trim/atrim+concat+atempo) -> mp4 download; ffmpeg graph validated locally
- rework.html: export panel (speed + speech-gap trim preview + SRT/video export),
client-side SRT from working segments, language selector (auto/ko/en/zh/ja)
- transcribeFromFile forwards optional language (Whisper auto-detect misfired -> zh)
Spec updated with the audio-silence -> speech-gap design correction.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
File-upload + local faster-whisper synced transcription, segment editor,
SRT export, with speed/silence-removal phases. N150 CPU, small int8 model.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Shared-CSS-first strategy (variables.css + style.css) to lift all 12
Thymeleaf pages, then per-page polish + light interactions. Minimal /
data-density direction; no backend or new libraries.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
