I ve added a Japanese beta to SubtitleGenerator for creators working with Japanese videos.
The update brings the generate edit style export workflow into a Japanese interface, with changes to Japanese caption joining and segmentation. It uses the same editor and plans as the English product.
Your first video is free without signing up. Free exports are 720p with a small watermark; subtitle files, watermark-free HD and translation are on Pro / Max.
If you make videos in Japanese, what still takes the most work after the first subtitle draft? I d love a short example of a caption or interface phrase you d change.
Hey Product Hunt — I built SubtitleGenerator to take a video from AI transcription to final export without bouncing between tools.
In one browser editor, you can fix text and timing, choose from 30 styles, translate a full track while keeping the original, and export video or subtitle files. The part I designed differently is Fix: uncertain source words are flagged, the remaining review count stays visible, and the transcript has a real finish line—“All clear.”
You can try it on one video without signing up. The Free workflow includes all 30 styles and 720p export with a small watermark. Pro and Max add high-accuracy transcription, full-track translation, HD without a watermark, eight subtitle formats, custom fonts, and saved brand styles. PAYG is there if you need more subtitle time but do not want another subscription.
Your original video stays in the browser. Only extracted audio is sent for transcription, and that audio is deleted after transcription is complete.
What still slows you down most after an AI transcript is generated? I’ll be here throughout launch day and would love blunt feedback on the Fix flow.
@woshiliyana Great Launch Yan!
Was the audio-only-upload thing a privacy ask from users, or just how you'd build it anyway?
@yashekbote Thanks, Yash! A bit of both. Keeping the original video in the browser reduces the privacy concern and avoids a large video upload. Transcription only needs the audio, so sending the smaller extracted track felt like the better default. I also delete it after transcription is complete.
Streamlining video to publish-ready captions straight in the browser saves creators so much time. Super cool launch, congrats!
@thisiskp_ Thanks, KP! I’m a YouTuber myself and work with subtitles often, so SubtitleGenerator grew directly out of my own workflow and frustrations. The same is true of TimedSubs—both products started with problems I wanted to solve for myself. Really appreciate the support!
Hey everyone! I’ve opened the official SubtitleGenerator Discord for support, updates, and feedback. I’d love to see you there: https://discord.gg/ey8sEqwtXR
the honest answer for me is speaker labeling when two people talk over each other or overlap slightly. the transcript reads fine as text but figuring out which line belongs to who post-transcript, especially in an interview or podcast clip, always ends up as manual work after the "AI part" is technically done. is that on your radar at all or is this more aimed at single-speaker content like tutorials/voiceovers for now
@galdayan It’s definitely on my radar. In an early two-speaker test, all four speaker turns were identified correctly, with confidence scores between 0.94 and 1.00, so the results look promising. Speaker labels and manual corrections are planned for v2. Overlapping speech is the harder case, and I want to test it on real interviews and podcasts before making any promises. Thanks for bringing this up!
@woshiliyana 0.94-1.00 on a clean two-speaker test is a good sign honestly, that's higher confidence than I expected for turn-taking detection. makes sense to hold off on promises for overlap until you've thrown real interviews at it, that's usually where the messy cross-talk actually lives, not the clean back-and-forth. will keep an eye out for the v2 labeling
@galdayan Thanks, Gal! Stay tuned—I’ll share an update once the feature reaches the quality bar I’m aiming for. You’re very welcome to try SubtitleGenerator in the meantime; I’ve made the free allowance quite generous, so there’s plenty of room to experiment.