imansur-ytb 0.3.0

A versatile tool for YouTube media downloading and AI processing with SQLite-based glossary.

imansur-ytb

A professional, modular AI pipeline for high-quality YouTube media ingestion and technical translation.

Philosophy: Modular Pipeline Architecture

This library is designed as a series of independent pipeline components. Each module is decoupled, meaning you can use the downloader without the translator, or provide your own transcripts to the translation engine.

Installation

pip install imansur-ytb

Module 1: Media (The Collector)

Handles media downloading with zero-touch FFmpeg management.

from imansur_ytb import get_video

# Returns a list of metadata dictionaries containing file paths
media_list = get_video("https://youtube.com/watch?v=...", audio_only=True)

Module 2: Translator (The Intelligence)

A standalone technical translation engine optimized for English-to-Turkish dubbing scripts. It features AI Term Discovery and persistent SQLite memory.

from imansur_ytb import translate_transcript

# Use as a standalone tool for any transcript JSON
translated_data = translate_transcript(
    transcript_path="my_transcript.json",
    api_key="YOUR_GEMINI_KEY", # or set GEMINI_API_KEY in .env
    verbose=True
)

Key Features of Module 2:

• Zero-Config .env Support: Automatically picks up GEMINI_API_KEY.
• SQLite Glossary: Ships with a default 1300+ term technical dictionary.
• Auto-Learning: Discovers new terms during translation and updates the local glossary.
• Consistency: Uses a double-pass AI logic to ensure terminology remains consistent across the entire transcript.

Parameter Reference

translate_transcript(...)

Parameter	Type	Default	Description
transcript_path	str	Required	Path to the input JSON transcript
api_key	str	None	Gemini API Key (Will check .env if None)
model	str	"gemini-2.0-flash"	AI Model (Optimized for Flash)

Module 3: Synthesis (The Voice)

Converts text or transcripts to high-quality neural speech using edge-tts.

from imansur_ytb import synthesize_text, synthesize_transcript

# 1. Simple TTS (for PBX/Switchboard)
synthesize_text("Hoş geldiniz!", "welcome.mp3", voice="tr-TR-AhmetNeural")

# 2. Pipeline Dubbing (Timed & Synced)
synthesize_transcript("tr_transcript.json", output_file="full_dub.mp3", verbose=True)

Key Features of Module 3:

• Free & Unlimited: No API tokens required (uses Microsoft Edge TTS).
• Auto-Sync: Automatically calculates if a segment needs to be sped up to fit the video timing.
• Timed Merging: Uses FFmpeg to place audio segments at exact timestamps with correct silences.

---

Developed by imansur - Building the future of AI Dubbing.