Subtitles Repository

This repository is dedicated to preparing subtitles as part of working with foreign languages and testing Kardenwort tools. Check the Release Notes for the latest updates.

Table of Contents

• Subtitles Repository
  • Table of Contents
  • Directory Overview
  • Subtitle Creation Process
    • 1. Acquisition
    • 2. Slicing (Optional)
    • 3. Recognition
    • 4. Version Control
    • 5. Correction
    • 6. Sentence Merging
    • 7. Translation
    • 8. Verification
    • 9. YouTube Studio Upload
    • 10. Learning Integration (Anki)
  • Notes
  • Kardenwort Ecosystem
  • License

Directory Overview

• 20251223140658-lab: Test records.
• 20251228104519-p-aia-ov: Podcast content https://podcast.onvibe.io/

Return to Top

Subtitle Creation Process

The following optimal process has been established for creating subtitles, leveraging Subtitle Edit and a modified Whisper utility (supporting word-by-word mode).

1. Acquisition

Receive video or audio recordings (mp3/mp4) using tools like Open-Video-Downloader or NewPipe. Low resolution is sufficient for this purpose.

2. Slicing (Optional)

For very long recordings, use LosslessCut to split the file along Chapter boundaries into smaller portions for parallel or piece-wise processing.

3. Recognition

Generate text from the audio track.

• Tools: Subtitle Edit or Whisper Audio Recorder & Transcriber (Faster-Whisper, recommended model: large-v3-turbo).
• Settings: --threads 8 --one_word 2
• Output: An SRT file with word-by-word subtitles (initial v1), offering maximum flexibility.

4. Version Control

• Create a separate directory in the Git repository.
• Save the subtitles and track changes.
• Note: Ensure only text subtitle files are tracked; exclude binary files.

5. Correction

• Open the subtitles in Subtitle Edit.
• Play the media and edit using Ctrl+H and "Merge line".
• Commit changes to Git after each session (include ZID).
• Result: Validated word-by-word subtitles (v1).

6. Sentence Merging

• In Subtitle Edit, use Auto-translate / Merge sentences....
• Merge words into sentences based on audio track boundaries.
• Result: Merged phrase/sentence subtitles (v2). This version prepares the file for translation.

7. Translation

• Use Subtitle Edit's Auto-translate / Auto-translate via Copy-Paste....
• Service: aistudio.google.com (Gemini 2.5 Pro).
• Prompts:
  • English: docs/prompts/20251108170321-translation-into-english.md
  • Belarusian: docs/prompts/20251108170321-translation-into-belarusian.md
• Settings: Max block size 5000-10000, empty model context.
• Format: Line separator @@@.
• Output: Save with .en postfix (e.g., filename.v2.en.srt).

8. Verification

• Integrity: Create a hash sum of the audio file (using 7z) to ensure subtitle synchronization.
• Review: View double subtitles in a suitable player (e.g., Chrome with asbplayer plugin, or PotPlayer on Windows).

9. YouTube Studio Upload

Finalize the process by making the subtitles available on YouTube.

• Access: Go to YouTube Studio.
• Subtitles Section: Select the video and navigate to the "Subtitles" tab.
• Add Language: Add the target language (e.g., English or Russian).
• Upload File: Choose "Upload file" and select the v2 (or translated .en) SRT file.
• Sync: Choose "With timing" to ensure the internal SRT timestamps are used.
• Publish: Review and publish to make them live.

10. Learning Integration (Anki)

The finalized v2 subtitles can be used for language learning.

• Deck Preparation: Use the subtitle text to break down and prepare an Anki deck using Kardenwort.
• Word Processing: Process each word using the IntelliFiller AI Anki add-on to automatically fill in card details and context.

Return to Top

Notes

The aistudio.google.com service is free to use, but please be aware of the following:

• There are usage limits (rate limits) for the free tier.
• By using the free tier, you agree that Google may use the data entered into the chat for service improvement purposes.

Return to Top

For more details, see:

• RFC 20251228230424 - Release v1.0.12 (Belarusian Subtitles)
• RFC 20251228193409 - Subtitle Creation Process

Kardenwort Ecosystem

This project is part of the Kardenwort environment, designed to create a focused and efficient learning ecosystem.

Return to Top

License

This project is licensed under the MIT License.

See the LICENSE file for details.

Return to Top