Subtitle Generator AI

Free offline AI subtitle generator for Windows, macOS and Linux powered by OpenAI Whisper

Turn any video into accurate subtitles (SRT) or burned-in captions with one click.
100% local: no cloud, no account, no upload, no subscription.

_{Batch subtitling five videos on a laptop CPU everything runs on your own machine.}

---

Subtitle Generator AI is a free desktop app that automatically creates subtitles for video files using OpenAI Whisper speech recognition, running entirely on your own computer through whisper.cpp. Drop in videos, press Start, and get time-synced .srt subtitle files plus, if you want, a copy of each video with the subtitles burned in. It supports 99 spoken languages, processes whole folders in one batch, and works offline after a one-time model download. Made by Inbora Studio.

Contents

• Why Subtitle Generator AI
• Download
• How to generate subtitles for a video
• Features
• Screenshots
• Choosing a Whisper model
• FAQ
• How it works
• Build from source
• Roadmap
• Community and support

Why Subtitle Generator AI

	Subtitle Generator AI	Typical online subtitle tools
Privacy	Video never leaves your PC	Video is uploaded to a server
Cost	Free, unlimited minutes	Often per-minute pricing or monthly plans
Internet	Only for the one-time model download	Required for every video
Batch processing	Whole folders, parallel jobs	Usually one file at a time
Burned-in captions	Built in, with live style preview	Often a separate paid step
Accuracy	Choose Whisper Tiny → Large v3	Fixed by the service

Ideal for YouTubers, podcasters, educators, students, journalists, translators, accessibility work and anyone who needs captions for sensitive or unpublished footage that must not be uploaded.

Download

Platform	File	Notes
Windows 10 / 11 (64-bit)	SubtitleGeneratorAI-Setup-<version>.exe	One-click installer. No admin rights needed.
Windows (no install)	SubtitleGeneratorAI-Windows-<version>-portable.zip	Unzip anywhere and run SubtitleGeneratorAI.exe.
macOS (Apple Silicon / Intel)	.dmg	Metal GPU acceleration on Apple Silicon.
Linux (x86-64)	.AppImage	chmod +x and run.

Nothing else to install. The Windows installer bundles the Whisper engine, FFmpeg and every runtime library, so there are no "missing DLL" errors and no FFmpeg setup. On first launch the app offers to download a Whisper model sized for your PC; after that it works fully offline.

How to generate subtitles for a video in 3 steps

1. Install and open Subtitle Generator AI. On first launch, click Choose a Model and download the model the Model Manager recommends for your PC (one-time).
2. Add videos drag and drop files or whole folders onto the window, or use Add Files / Add Folder. You can also drop videos straight onto the app's desktop icon.
3. Press Start Processing. Each video gets a .srt subtitle file in your output folder, and (optionally) a <name>_subbed.mp4 copy with the subtitles burned in.

That's it open the .srt in any video player (VLC, MPC-HC, mpv, Windows Media Player), upload it to YouTube, or import it into Premiere Pro, DaVinci Resolve or CapCut.

Features

Transcription

• Automatic speech-to-text with OpenAI Whisper, from Tiny (fast) to Large v3 (most accurate)
• 99 spoken languages with automatic language detection
• Fast mode (greedy decoding) or accurate mode (beam search)
• GPU acceleration: NVIDIA (CUDA), AMD (ROCm/HIP or Vulkan), Intel (Vulkan), Apple Silicon (Metal) or CPU only

Workflow

• Batch processing: queue hundreds of videos, scan folders recursively, run several jobs in parallel
• Pause, resume and cancel a batch; live progress, elapsed time and ETA per batch
• Works with MP4, MKV, AVI, MOV, WEBM, FLV, WMV, M4V, TS, MTS, M2TS, MPG, MPEG
• Model Manager recommends the best Whisper model for your RAM and downloads it in one click

Output

• SRT subtitle files with accurate timestamps, ready for YouTube and any editor
• Burned-in (hardcoded) subtitles exported as MP4 (H.264 + AAC)
• Subtitle styling font, size, colour, outline and position with a live preview
• Optional translation into 24 languages through a self-hosted LibreTranslate server

Desktop app

• 100% offline and private no telemetry, no account, no cloud
• Clean dark interface with live CPU / RAM / GPU usage and a filterable log console
• Remembers all settings, window size and position between sessions
• Keyboard shortcuts for everything (see FAQ)

Screenshots

Video queue — drop files or whole folders.	Settings — model, subtitle style with live preview, export and translation.
Model Manager — recommends a Whisper model for your RAM.	Done — open the output folder straight from each row.

Burned-in subtitles produced by the app (white text, black outline, bottom centre):

The generated .srt file for that video:

1
00:00:00,000 --> 00:00:04,020
Good morning everyone, and thank you for joining us.

2
00:00:04,020 --> 00:00:11,460
Today we are introducing a faster, simpler way to create subtitles for every video you publish.

3
00:00:11,460 --> 00:00:16,000
It works offline, it is free, and it takes a single click.

Choosing a Whisper model

Download models from the in-app Model Manager; it highlights the recommended one for your PC. Bigger models are more accurate but slower and need more memory.

Model	Download	RAM needed	Languages	Best for
Tiny / Tiny (English)	75 MB	~0.5 GB	99 / English	Quick drafts, very old PCs
Base / Base (English)	142 MB	~1 GB	99 / English	Fast everyday English captions
Small / Small (English)	466 MB	~2 GB	99 / English	Good balance of speed and accuracy
Medium / Medium (English)	1.5 GB	~5 GB	99 / English	High accuracy, accents, noisy audio
Large v2	2.8 GB	~10 GB	99	Very high accuracy
Large v3	2.8 GB	~10 GB	99	Best accuracy, especially non-English

Tip: the (English) models are faster and slightly more accurate for English-only videos. For any other language, use a multilingual model.

Models are stored outside the app, so updates and reinstalls keep them: %LOCALAPPDATA%\SubtitleGeneratorAI\models (Windows), ~/Library/Application Support/SubtitleGeneratorAI/models (macOS), ~/.local/share/SubtitleGeneratorAI/models (Linux).

FAQ

How do I generate subtitles for a video for free, without uploading it?

Install Subtitle Generator AI, download a Whisper model once from the Model Manager, add your videos and press Start Processing. Transcription runs locally with whisper.cpp, so the video never leaves your computer and there is no usage limit or fee.

Does it work offline?

Yes. Internet is only used to download a Whisper model the first time (and optionally to check for updates). After that, transcription, subtitle export and burning captions into video all work with no connection.

Is it really free?

Yes. There are no accounts, subscriptions, watermarks or per-minute limits.

Which languages does it support?

Whisper recognises 99 spoken languages, including English, Spanish, French, German, Hindi, Arabic, Chinese, Japanese, Korean, Portuguese, Russian, Indonesian and Turkish. Leave the language on Auto Detect or pick it in Settings. Optional translation of the subtitles into 24 languages is available through a self-hosted LibreTranslate server.

What subtitle formats does it create?

It writes standard SubRip (.srt) files, which work with YouTube, VLC, Premiere Pro, DaVinci Resolve, Final Cut (via import), CapCut and almost every player. It can also export an MP4 with hardcoded (burned-in) subtitles for social media, where soft subtitles are often not supported.

How accurate is it?

Accuracy depends on the model and the audio. Large v3 is very accurate on clear speech; Small and Medium are a good balance on most PCs; Tiny and Base are fast but make more mistakes with names, accents and background noise. You can always edit the .srt in any text editor.

How fast is it? Do I need a GPU?

No GPU is required. On a modern CPU, small models transcribe faster than real time; the demo above (five short clips, Tiny model) finished in 13 seconds including burning in the subtitles. A GPU (NVIDIA, AMD, Intel or Apple Silicon) makes the large models much faster.

Can I subtitle many videos at once?

Yes. Add files or whole folders (subfolders are scanned too), then press Start. Increase Parallel Jobs in Settings to process several videos at the same time if you have enough RAM. You can pause, resume or cancel the batch at any time.

Can I change how burned-in subtitles look?

Yes. In Settings → Subtitle Appearance choose the font, size, colour, outline thickness and position (nine positions). The preview next to the controls shows exactly how the captions will be rendered.

Is it a good alternative to paid or cloud subtitle services?

For most videos, yes: it runs OpenAI's open Whisper speech-recognition models locally, for free and without limits, and keeps your footage private. Cloud tools can still be handy for collaborative editing in a browser.

Are there keyboard shortcuts?

Shortcut	Action
Ctrl+O / Ctrl+Shift+O	Add files / add a folder
Ctrl+Enter	Start processing
Ctrl+1 / Ctrl+2 / Ctrl+3	Input / Settings / Progress page
Ctrl+M	Open the Model Manager
Ctrl+`	Show or hide the log console
Delete	Remove selected files from the queue
F11	Maximise / restore the window

Where are my settings and logs?

Settings are saved automatically (Windows registry under HKCU\Software\SubtitleGeneratorAI). Logs are in %LOCALAPPDATA%\SubtitleGeneratorAI\SubtitleGeneratorAI\logs on Windows; you can also view, filter, copy and save them from the in-app Console.

Windows SmartScreen warns about the installer. Is it safe?

New, unsigned apps can trigger SmartScreen until they build up reputation. Click More info → Run anyway. You can verify the download against SHA256SUMS.txt attached to every release, or build the app yourself from this repository.

How it works

flowchart LR
    A[Video file<br/>MP4 MKV MOV ...] -->|FFmpeg| B[16 kHz mono audio]
    B -->|whisper.cpp<br/>Whisper model| C[Timed transcript]
    C -->|optional| D[LibreTranslate<br/>translation]
    C --> E[.srt subtitles]
    D --> E
    E -->|FFmpeg + libass<br/>optional| F[MP4 with<br/>burned-in subtitles]

Loading

• Audio extraction: FFmpeg decodes the video's audio track to 16 kHz mono PCM.
• Speech recognition: whisper.cpp, a C/C++ port of OpenAI Whisper, runs the selected model on the CPU or GPU (via ggml).
• Subtitles: segments with start/end timestamps are written as SubRip .srt.
• Burn-in: FFmpeg's subtitles filter (libass) renders the captions with your style and re-encodes the video as H.264/AAC MP4.
• App: native C++17 with Qt 6 a small, fast desktop app with no Python, Electron or browser runtime.

Build from source

Requirements

• CMake 3.25+
• A C++17 compiler: GCC 13+, Clang 15+ or MSVC 2022
• Qt 6.8+ (Widgets, Core, Concurrent, Network)
• FFmpeg at runtime (bundled automatically in Windows release builds)
• whisper.cpp is pinned as third_party/whisper.cpp; release scripts fetch the exact commit

git clone https://github.com/InboraStudio/Subtitle-Generator-AI.git
cd Subtitle-Generator-AI

Windows (MSYS2 UCRT64) one-step installer build

Install MSYS2 and Inno Setup 6, then in a UCRT64 shell:

pacman -S mingw-w64-ucrt-x86_64-toolchain mingw-w64-ucrt-x86_64-cmake \
          mingw-w64-ucrt-x86_64-ninja mingw-w64-ucrt-x86_64-qt6-base \
          mingw-w64-ucrt-x86_64-qt6-tools

Double-click build_installer.bat (or run bash scripts/package-windows.sh). It fetches the pinned whisper.cpp, builds a release, bundles Qt, the runtime and FFmpeg, verifies that no DLL is missing on a clean PC, and writes the installer and portable zip to dist\.

For a quick development build instead:

cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release
cmake --build build -j
./deploy.sh            # copy Qt + runtime DLLs next to the exe

"The procedure entry point qResourceFeatureZstd could not be located" means a different Qt6Core.dll was picked up from PATH. Run ./deploy.sh so the exe ships with its own matching DLLs.

Linux

# Arch / CachyOS
sudo pacman -S base-devel cmake ninja qt6-base ffmpeg
# Debian / Ubuntu
sudo apt install build-essential cmake ninja-build qt6-base-dev ffmpeg
# Fedora
sudo dnf install gcc-c++ cmake ninja-build qt6-qtbase-devel ffmpeg

cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release
cmake --build build -j
./build/bin/SubtitleGeneratorAI

For a Vulkan build, also install the Vulkan headers, loader, SPIR-V headers and glslc (e.g. vulkan-headers spirv-headers shaderc vulkan-icd-loader on Arch, libvulkan-dev spirv-headers glslc on Debian/Ubuntu) and add -DGPU_BACKEND=vulkan.

macOS (Apple Silicon or Intel)

make deps      # brew install cmake ninja qt ffmpeg
make           # fetches the pinned whisper.cpp, configures and builds
make run

The Makefile builds for the machine's native architecture (arm64 on Apple Silicon, even from a Rosetta terminal). make doctor prints the detected toolchain. Metal GPU acceleration is enabled automatically.

GPU acceleration

Inference runs on the CPU by default. Choose a GPU backend when configuring:

-DGPU_BACKEND=	Hardware	Extra build dependencies
none (default)	CPU only	—
vulkan	AMD, Intel, NVIDIA	Vulkan headers + loader, glslc, SPIR-V headers
cuda	NVIDIA	CUDA Toolkit
hip	AMD (ROCm)	ROCm / HIP toolchain

On macOS, Metal is enabled automatically leave GPU_BACKEND at none. Vulkan is the recommended choice for AMD and Intel GPUs: no vendor SDK, and recent hardware gets cooperative-matrix (tensor core) kernels through ggml.

Making a release

Push a version tag and GitHub Actions builds the Windows installer, portable zip, Linux AppImage and macOS DMGs, then publishes them with checksums:

git tag v2.1.0
git push origin v2.1.0

Roadmap

• whisper.cpp integration with GPU backends (Vulkan, CUDA, ROCm, Metal)
• Batch queue with parallel jobs, pause / resume / cancel
• Burned-in subtitles with live style preview
• One-click Windows installer with everything bundled
• More export formats (VTT, ASS) and a built-in subtitle editor
• Real-time microphone transcription
• Speaker diarization (who said what)

Have an idea? Open an issue or tell us on Discord.

Community and support

• Discord: discord.gg/pVA8NZ5mru help, feedback and new releases
• Bugs and feature requests: GitHub Issues
• Like the project? Star the repo it helps other people find a free, private subtitle generator.

Credits

Made by Inbora Studio (Dr Chamyoung).

Built on the shoulders of OpenAI Whisper, whisper.cpp / ggml, Qt and FFmpeg. Windows builds include FFmpeg from BtbN/FFmpeg-Builds (GPL); its licence ships in the app's bin folder.

_{Subtitle Generator AI free offline AI subtitle generator · automatic captions · speech to text · SRT generator · Whisper desktop app for Windows, macOS and Linux}