SubtitleYC 0.6.0

SubtitleYC pulls the captions that are baked into a video's picture back out as a proper subtitle file you can edit, retime and reuse.

It is free and open source under the MIT licence, it runs on Windows, and every bit of the text recognition happens on your own PC rather than on a server.

Version 0.6.0 swaps the old multi-gigabyte downloads for a 2.4 MB web setup that fetches the app and the OCR engine for you and checks both before installing.

What SubtitleYC Actually Does

Most subtitles are a separate text track riding alongside the video, and any subtitle editor can open them. Burned-in subtitles are different: the words were drawn onto the picture when the video was made, so to the computer they are just pixels, no different from a face or a wall.

That is why Subtitle Edit shows no subtitle track when you open one of these files. There is nothing to find.

SubtitleYC reads the pixels instead. You drag a box over the caption area, pick the language, and it runs OCR - optical character recognition, the same idea as scanning a paper letter into editable text - across every frame, then groups what it reads into timed subtitle lines.

The recognition engine underneath is VideOCR 1.5.1, built on PaddleOCR, and SubtitleYC wraps it in a normal desktop window so you never touch a command line. The result is saved as SRT, ASS or plain TXT.

The Editor Is the Half That Matters

No OCR is perfect, and reading text off compressed video with moving backgrounds behind it is about as hard as OCR gets. Expect a few wrong letters and a few lines that start or end a frame early.

SubtitleYC is built around fixing that. The editor shows every line next to the video, you can step one frame at a time with the arrow keys, nudge a line's start or end by a chosen number of frames, shift the whole file earlier or later, and snap every line to the video's frame grid.

You can also add and delete lines, apply bold, italic, underline or a colour to part of a line, undo and redo, and jump between subtitle boundaries with Shift and the arrow keys. Existing SRT, ASS or SSA files can be loaded into the same editor, so it doubles as a light subtitle editor even when no OCR is involved.

Getting the Video In

Open a local file, or paste a video URL and SubtitleYC downloads it with a bundled copy of yt-dlp, checks the available formats, and lets you pick one and a save folder.

If the site publishes its own captions or auto-captions, SubtitleYC can fetch those too and attach one to the project, which is worth trying before you run OCR at all. FFmpeg is bundled as well, for reading the video and pulling preview frames.

The URL download is a convenience, not the point of the app. If downloading is all you want, YTDLP-Interface is the lighter tool, and either way only download video you have the right to keep.

The 2.4 MB Installer and What It Fetches

The download on this page is a small web setup, so you need an internet connection while it runs.

• Run the setup EXE. It downloads the app and the CPU OCR engine from the developer's release page, about 0.9 GB together, and verifies both against their SHA-256 checksums.
• Let it install to Program Files and create the shortcuts. Your projects, settings and logs live separately under your local app data folder.
• Launch SubtitleYC. CPU OCR is ready on first run - nothing else to set up.

The three ZIP files on the developer's release page are payloads the installer fetches, not separate downloads. There is no standalone offline installer for 0.6.0.

CPU or GPU OCR?

CPU OCR works on any PC and is installed by default. It is slower, which matters on a two-hour film, but it needs nothing extra.

GPU OCR is optional and Nvidia-only: a supported card from the GTX 16 series through the RTX 50 series with a driver that supports CUDA 12.9. It is a 1.6 GB download from inside the app, with pause and resume, and once it lands you switch on GPU Acceleration in the OCR settings.

Not sure what card you have? GPU-Z will tell you in one window, and a current NVIDIA Studio driver is the safe choice if the app reports the driver as too old.

Which Languages It Reads

The first-run setting is English plus Simplified Chinese, and you can switch to a single language for cleaner results. The bundled models cover Simplified and Traditional Chinese, Japanese, Korean, Vietnamese, Thai, Indonesian, Malay, Filipino, Hindi, Marathi, Nepali, Tamil, Telugu, Arabic, Persian, Urdu, Uyghur, Turkish, Kazakh and Mongolian, along with the major European languages.

The interface itself is in English or Simplified Chinese, and the installer's language choice carries into the app.

Who Should Look Elsewhere

The tell is a subtitle track. Open the file in MKVToolNix and look at the track list: if a subtitle track is there, your subtitles are not burned in, and extracting that track is faster and perfectly accurate.

Our guide to remuxing without re-encoding covers that route, and Subtitle Edit opens the extracted track directly.

If there are no subtitles at all and you want them generated from the speech, that is a different job, and AutoSubs does it. TV recordings with closed captions are CCExtractor territory.

DVD and Blu-ray subtitles are bitmap images rather than burned-in video, and SubRip is the long-standing tool for OCR on those. The older AviSubDetector tackles the same burned-in problem as SubtitleYC but has not moved past beta.

If you already have a subtitle file and only the timing is off, a dedicated timing editor is quicker: Subtitle Workshop or VisualSubSync for full editing.

For a plain shift of every line by the same amount, Time Adjuster does that and nothing else.

Tero Subtitler is another maintained editor worth a look if you want styling tools SubtitleYC does not have. And this is Windows-only software, with no macOS or Linux build.

Finally, the developer calls 0.6.0 an early release and says so plainly: keep a copy of anything important and read the subtitles it produces before you rely on them.

Using the Subtitles Afterwards

Save the SRT next to the video with the same filename and VLC Media Player or MPC-HC will pick it up automatically.

For a player that does not, MPC SubtitleSource makes external subtitle files loadable in any DirectShow player. To bake the file into an MKV as a proper track, MKVToolNix does it in seconds without touching the video.

If you want a version with the old burned-in captions gone and the new clean track in their place, that is a re-encode with a crop, and HandBrake is the tool.

What You Need

• 64-bit Windows 10 or 11. The developer does not state a floor; the Qt 6 framework underneath the app requires it.
• About 1 GB of disk space for the app and CPU OCR, plus 1.6 GB more if you add GPU OCR.
• An internet connection during setup, because the installer downloads the app and engine.
• For GPU OCR only: an Nvidia GTX 16 through RTX 50 card with a CUDA 12.9-capable driver.

What Is New in 0.6.0

• New 2.4 MB web installer that downloads, verifies and installs the app and CPU OCR in one pass.
• Optional Nvidia GPU OCR with driver checks, download progress, pause and resume, and checksum verification.
• Clearer reporting of download size, installed size and temporary space, and the app now distinguishes its own version from the VideOCR engine version.
• Smoother video scrubbing, with rapid slider moves batched and the final position preserved.
• Fixes for the native video preview covering dialogs, plus protection against losing unsaved subtitle edits.

Quick questions

The rest of the Subtitle Tools category covers editing, syncing and converting once you have a subtitle file in hand.

Download SubtitleYC 0.6.0 and see how it handles your files, or tell us how you got on if you have used it before.