Offline Subtitle Generators in 2026: 5 Tools That Run on Your Machine

An offline subtitle generator runs the speech model on your own computer, so the video is never uploaded and there is no length or file-size limit. In 2026 almost all of them are built on OpenAI's Whisper, and the real differences are in what happens after transcription: editing, translation, styling and burn-in. The five tools below cover the free, the Mac-only, the do-it-yourself and the multilingual end of that range.

We make one of them. Sablate is in this list, and so are four tools that do some jobs better than it does. Each section says which.

Why generate subtitles offline?

Three reasons come up again and again, and only one of them is privacy.

1. Length and size limits disappear. Browser tools upload the whole file first, which is why they cap video length and file size. A two-hour lecture or a 20 GB camera file is ordinary work for a local app.
2. The footage stays with you. An unreleased cut, a client's recording, a medical or legal interview — anything under an NDA or a data-protection policy is simpler when there is no third-party server involved at all.
3. The cost does not scale with minutes. Cloud transcription is billed per minute or per plan. A local model costs the same whether you transcribe one hour a month or a hundred.

The trade-off is your hardware. A local model is as fast as your machine, and the most accurate Whisper models want a decent GPU to feel quick. The Whisper model comparison shows what each model size costs in time and memory.

What to look for in an offline subtitle generator

1. Which Whisper models it runs. The smallest models are fast and rough; large-v3 and large-v3-turbo are where accuracy becomes publishable with light correction.
2. GPU support. NVIDIA on Windows and Apple Silicon on Mac are what make the larger models practical.
3. A real editor. You will correct the output. A timeline with draggable timings saves more time than any accuracy difference between tools.
4. Export formats. .srt for everything, .vtt for the web, .ass if the styling has to survive.
5. Burn-in. TikTok, Reels and Shorts accept no subtitle files, so for social video the captions have to be rendered into the picture.
6. Translation. Whisper itself can only translate into English. Anything else needs a separate translation step.

The 5 best offline subtitle generators in 2026

1. Subtitle Edit — the free standard for subtitle files

Subtitle Edit is free, open source, and the tool professional subtitlers already use. It has Whisper transcription built in, with several engines to choose from, reads over 300 subtitle formats, and has the deepest set of timing, conformance and quality-check tools on this list.

It can also translate through several engines and burn subtitles in through an FFmpeg-based export. What it does not try to be is a caption designer: styled social captions and per-language fonts are not what it is built for, and the interface assumes you know subtitling vocabulary.

Free. If your deliverable is a correct subtitle file, start here.

2. Buzz — the simplest free transcriber

Buzz is a free, open-source desktop app for Windows, macOS and Linux that transcribes audio and video offline with Whisper, whisper.cpp or faster-whisper. You drop a file in, pick a model, and export .txt, .srt or .vtt. It also has a watch folder that transcribes new files automatically, and a command-line mode.

It is a transcriber rather than a subtitle editor: there is no timeline, no styling and no burn-in. Its translation is Whisper's own, which means into English only. Recent Mac versions require Apple Silicon.

Free. The right pick when you want a transcript or a first-draft .srt with the least setup.

3. MacWhisper — the most polished on a Mac

MacWhisper is a native macOS transcription app with a clean interface and good performance on Apple Silicon. The free tier runs the smaller models; Pro adds the larger ones, batch transcription, .srt and .vtt export and speaker labels.

It is a transcription tool first. It does not burn subtitles into video, so social captions need a second tool, and it does not run on Windows at all.

Free tier; Pro is €59 one-time as of September 2026. The best choice for Mac users who mainly want accurate transcripts and subtitle files.

4. Sablate — styled burn-in and many languages from one job

Sablate is a desktop app for Windows and macOS built for the part after transcription. It runs faster-whisper locally with four model tiers — Fast, Balanced, Accurate and Best — and on an NVIDIA GPU it is roughly five to ten times faster than the same machine on CPU.

The difference from the rest of this list is that translation is a layer on the same job rather than a new one. Transcribe once, add Spanish, Arabic and Japanese, and each layer keeps the original timing while carrying its own style and a font that can actually draw its script. Translation runs offline through Argos, or through DeepL, OpenAI or Claude with your own API key.

The editor has timeline retiming, split and merge, a reading-speed warning and find and replace, and thirteen caption presets from broadcast-style Classic to word-by-word Hormozi-style captions. Export covers .srt, .vtt, .txt, .ass, burned-in .mp4 at 16:9, 9:16 or 1:1 rendered with libass, and a single .mkv with one subtitle track per language.

Free for videos up to ten minutes; Pro is $29 once. Free includes three jobs a month, the Fast and Balanced models, one offline translation layer and burn-in with a small corner mark. There is no Linux build, and no speaker labels.

5. Whisper and FFmpeg on the command line — full control, no interface

The tools above all wrap the same model. Running it directly — OpenAI's whisper command, faster-whisper or whisper.cpp — gives you every option they hide, works on any operating system, and costs nothing.

whisper interview.mp4 --model large-v3 --language en --output_format srt

FFmpeg then burns the result in if you need it. The price is that everything else — correcting text, fixing timings, styling, translating — is manual work in a text editor, and GPU setup on Windows is its own project.

Free. The right choice for scripting a batch pipeline, and the wrong one if you have to correct the output by hand.

Offline subtitle generators compared

Prices checked on 24 September 2026.

Four of those rows go against Sablate. Subtitle Edit and Buzz are free outright, open source and run on Linux; Sablate is none of those. MacWhisper labels speakers, which Sablate does not do. Subtitle Edit handles over 300 subtitle formats to Sablate's four. If any of those decide it for you, they decide it.

Which one should you pick?

"I need a correct .srt and I will fix it myself." Subtitle Edit. Free, and deeper than anything paid here for subtitle work.

"I just want a transcript, quickly." Buzz on any platform, or MacWhisper on a Mac.

"I am making captions for TikTok, Reels or Shorts." A tool that renders styled burn-in. Doing it with an .srt and FFmpeg works, but the styling becomes the job — the burn-in guide shows both routes.

"The same video has to go out in four languages." Sablate. One transcription with a styled layer per language is the case it was built for, and the multilingual workflow explains why that beats running the model once per language.

"I am scripting a pipeline for hundreds of files." The command line. No interface is faster than no interface.

The bottom line

Every tool here keeps your video on your machine, so the choice is about the output, not the transcription. For subtitle files, the free tools are excellent and you should use them. For styled captions and multilingual delivery, a free tool plus FFmpeg becomes a lot of manual work.

Sablate's free plan covers videos up to ten minutes with no account and no card. Download it for Windows or Mac and run one real video through it, or compare the plans first.