Glossary

AI & Transcription Terms

AI Model
The “brain” that converts speech to text. Larger models are more accurate but slower. Think of it like different quality cameras – a better camera captures more detail but takes longer to process.

Whisper
OpenAI’s speech-to-text AI technology. Powers the transcription in Subtitle Creator.

Whisper.cpp
A faster version of Whisper optimized for Mac computers. Uses your Mac’s graphics chip for speed.

Beam Size
How many different word possibilities the AI considers. Higher = more accurate but slower. Like proofreading: reading once is fast, reading five times catches more mistakes.

Temperature
Controls creativity vs. accuracy. 0 = stick to most likely words (best for subtitles). 1 = more creative guessing (not recommended).

Hallucination
When AI invents text that wasn’t spoken. Common examples: “[Music]”, “[Silence]”, or repeating the same line multiple times. Usually happens during silence or loud background noise.

Subtitle Terms

SRT File
Standard subtitle format (.srt extension). Contains timestamps and text. Works with all video players.

Timestamp
The exact time (hours:minutes:seconds) when each subtitle appears and disappears.

Time Offset
Shifts all subtitles earlier or later. Use this when subtitles don’t sync with speech.

Processing Terms

Hardware Acceleration
Using your Mac’s graphics chip (GPU) instead of just the processor (CPU) to work faster. Only works on Apple Silicon Macs (M1/M2/M3/M4).

Batch Processing
Processing multiple videos at once. Add several files and the app handles them one by one.

Apple Silicon
Apple’s own chips: M1, M2, M3, M4. Much faster than older Intel chips for AI work.

Quality Settings

Filter Non-Speech Audio
Removes music, sound effects, and background noise from transcription results.

Confidence Threshold
Minimum certainty required before AI adds a word. Higher = fewer guesses, more gaps.

Filter Repetitive Text
Removes stuttering or repeated phrases the AI might hallucinate.

Previous Context
Whether AI remembers what it just transcribed. Helps with continuous speech but can cause error cascades in movies.

File & Format Terms

MKV, MP4, MOV, AVI
Common video file formats. Like different ways to package the same movie.

FFmpeg
Software that extracts audio from video files. Required for the app to work.

Compression Ratio
Measures how repetitive text is. High repetition = likely hallucination.

Technical Terms

License Key
Your unique code from Gumroad that activates the app. Looks like: XXXX-XXXX-XXXX-XXXX.

Dependency
External software the app needs. Subtitle Creator needs FFmpeg and Whisper models.

Universal Binary
Single app file that works on both Apple Silicon and Intel Macs.

Application Support Folder
Hidden folder where macOS apps store their files: ~/Library/Application Support/SubtitleCreator/

Model Names Explained

Tiny – Smallest, fastest, least accurate (150MB)
Base – Small, fast, decent accuracy (142MB)
Small – Recommended for most users (466MB)
Medium – Slower but more accurate (1.5GB)
Large v2 – Very accurate, better for noisy scenes (3.1GB)
Large v3 – Newest, most accurate for clean audio (3.1GB)

The “.en” suffix means English-only model (slightly better for English than multilingual versions).

Translation Terms

Source Language
The language spoken in the video (currently English only).

Target Language
The language you want subtitles translated to (50+ options).

Gemini API
Google’s AI translation service. Free tier available.

Common Abbreviations

SRT – SubRip Text (subtitle format)
VAD – Voice Activity Detection
GPU – Graphics Processing Unit (graphics chip)
CPU – Central Processing Unit (main processor)
API – Application Programming Interface (how software talks to other software)
RAM – Random Access Memory (temporary working memory)