GenSRT v1.2 — Windows tool to auto-generate SRT/VTT subtitles using Whisper

[gsadasivan]

Hi everyone,

Sharing GenSRT, a free open-source Windows tool I've been building for
subtitle work. It generates SRT/VTT subtitles from video using OpenAI Whisper,
and gives you a full workflow in one window for any language Whisper supports.

What it does

• Drop a video, click Generate, get an SRT (and a VTT in the same place)
• Edit cues in a built-in player with live preview — split, merge, delete,
  edit text and timings
• Translate to 100+ languages via Google Translate (MyMemory fallback)
• Burn subtitles into a copy of the video with one click
• Bundled ffmpeg — no separate install needed
• SRT filenames are Plex / Jellyfin / Kodi-compatible

Short demo (generate -> edit -> translate -> playback):
https://www.youtube.com/watch?v=QnW1Mqwsltk

Where it goes beyond a basic Whisper front-end
v1.2 added proper support for fine-tuned models for under-served languages.
Community Indic fine-tunes (like vegam-whisper for Malayalam) previously
couldn't handle long-form audio — they'd transcribe a few seconds and stop,
because of a token-length limit in the Whisper decoder that's especially
punishing for Indic scripts. v1.2 works around it by slicing audio along
silences and transcribing each piece independently. If you've tried subtitling
Indian-language content and been let down, set the model dropdown to
smcproject/vegam-whisper-medium-ml-int8_float16 (auto-downloads on first
run). For everything else, large-v3-turbo is the sensible default.

Requirements

• Windows 10 or 11
• NVIDIA GPU recommended (any with ~2 GB VRAM) — much faster on GPU
• CPU mode also works (about 3x slower, same output). Tested on Intel Arc
  140V (Lunar Lake laptop).
• ~2 GB free disk for the model on first run

Download / source
GitHub: https://github.com/mountlord/GenSRT
Latest release - One click installer: https://github.com/mountlord/GenSRT/releases
License: AGPL-3.0 (free and open source)

Feedback welcome — what languages you tried, what worked, what didn't.

Thanks

[gsadasivan]

[This post has been deleted]

[gsadasivan]

GenSRT v1.2.5 released

Free, open-source subtitle generator for Windows. Transcribe → edit → translate → burn in, all in one window. Runs Whisper locally on your own machine — nothing is uploaded.

Download · Source (AGPL-3.0) · Demo video


Two installers now

• gensrt-install.exe — NVIDIA GPU (CUDA 12 driver) - 1,540 MB
• gensrt-install-cpu.exe — everything else. Much smaller download - 120 MB.

If you don't have an NVIDIA card, or you're on a metered connection, take the CPU build. It works on any x64 Windows machine.

One thing to know before you pick it: CPU transcription takes roughly nine times the length of your audio — a 5-minute clip is about 45 minutes on a 6-core desktop. Fine for short clips or an overnight batch. The CUDA build does the same clip in under three minutes.

The download got a lot smaller

GenSRT no longer bundles PyTorch. Transcription runs on CTranslate2 end to end, and PyTorch was only ever there to answer one question — "is a GPU present?" — which CTranslate2 answers itself. That was a large download every user paid for, including CPU-only users who could never benefit from it.

Translate to any language

Pick a target language in the Config panel, or [icode]--target-language ko[/icode] on the command line. Previously this only really worked for English: GenSRT batches subtitle lines into a single translation request separated by a marker, and that marker was itself getting translated — so with a non-English target most lines came back untranslated. Fixed.

Fixed in this release

• Subtitle text past two lines was being silently discarded on long lines
• Subtitles could overlap each other, which players handle inconsistently
• The window was clipped at 125% and 150% Windows display scaling — Burn SRT and Fullscreen fell off the right edge (thanks to the user who reported this)
• Forcing CPU mode in the config did nothing
• Burn-in failed silently on filenames containing brackets, commas or apostrophes — very common in scene releases
• Loading a language-variant subtitle file (clip.ml.srt) would quietly load clip.srt instead

Other changes

• Adding your own HuggingFace model now checks it's in the right format before downloading it, instead of failing later with an unhelpful error
• [icode]--self-check[/icode] verifies an installation is complete — useful if something isn't working
• WebVTT written alongside SRT, so output works directly in HTML5 video and Jellyfin
• Cue numbers shown in the subtitle list

Any language Whisper supports

The main advantage over plain Whisper is support for community fine-tuned models — you can plug in any CTranslate2 model from HuggingFace. That matters most for languages the stock Whisper models handle poorly. For Malayalam, the stock large-v3-turbo produced 19 mostly-garbage lines on a clip where a fine-tune produced 100 good ones. Similar fine-tunes exist for many Indic, Southeast Asian and African languages.

Known limitations — documented rather than hidden:

• Fine-tuned models emit occasional very short spurious cues (around 60 ms). Too brief to see during playback, but they're in the SRT. One-click cleanup is planned.
• A � at the end of a line means the model stopped mid-character. The text before it is fine.
• A single-language fine-tune will transcribe other languages phonetically rather than skipping them, so English commentary in a foreign-language broadcast comes out as nonsense in the target script.

Feedback and bug reports welcome — the DPI scaling fix in this release came from a user report.