Changelog

Fixed

• The dated output folder is created for you. If the folder you chose for exports does not exist yet, Substrike makes it before the first export instead of failing.
• Clearer wording in the watch box. The watch folder messages say what is happening and what to do.
• Recovery after a reload. If the window reloads, your batch and your place in it come back.
• Better layout at narrow widths. The side panel and the batch rows fit when the window is small.
• An unplugged drive is still caught. If the drive you save to goes missing, Substrike tells you.

Improved

• One button batch flow. Pick Today or Last session, or drop clips onto the window, and they are queued and processed in one go.
• Clearer pile rules. Most clips now land in Ready under Balanced, and the reason a clip is in Check is shown on its row.
• Unsure word counts match everywhere. The number you see in the batch, the editor and the summary is the same one, and rows say why words are unsure.
• New export presets. Share is now the default and makes smaller files that are easier to post. Discord and Original are there too.
• Voice chat dropdown. Setting up audio tracks is now one dropdown for where your friends are heard, instead of a list of presets.
• The watch folder follows OBS. It uses OBS's current recording folder, and it refuses to watch a whole drive.
• Montage size check. Substrike checks the size before it makes the reel, so you find out early. A montage for Discord is no longer re-encoded when it is auto-posted.
• Leave the Substrike ending off. There is a switch to turn the ending off.
• Translated rows. The no speech row, the batch summary line and the narrow side panel now read in Japanese too.

New

• Living captions style, with a Compare switch. Captions move with the voice, and Compare shows them next to the plain style so you can choose.
• Japanese interface (beta). Switch the app to Japanese in Settings. It is a beta, so tell us what reads oddly.

This release on GitHub

Fixed

• AI cleanup now works on installed copies. On a normal install the cleanup step could not find its helper and quietly did nothing. It now runs, and when it is not available the batch shows a calm banner instead of an error, with a fix button on every row.
• 240 fps and variable frame rate clips export with the outro. These clips could fail or lose the outro on export. They now come out whole.
• Smoother updates. Substrike updates in place, and no longer shows a false crash notice after an update.
• Your destination is remembered. It no longer resets after you visit the Style tab.
• The TikTok preview matches the export. What you see is where the captions land.
• The log shows real paths. Output and file paths in the log are the ones on your disk, and Play opens the video.

Improved

• Batch sorts clips into three piles. Ready clips go to your output folder. Clips with doubts go to a Check folder, each with the reason on its row. A clip with no speech writes nothing, and the row names the silent track.
• A new "When a clip has doubts" setting. Choose Careful, Balanced or Hands-off. Balanced is the default. Many clips with friends talking over each other land in Check under Balanced, and we will tune that in 1.0.5.
• Looks good moves a clip to Ready. After you review a Check clip, one click moves it over.
• Discord auto-post is stricter. Only clips that pass the strict bar are posted on their own.
• Fix buttons on every batch row, and a Stop batch button at the top.
• Proximity chat games. When your mic track is silent, captions can come from the party track instead. Two new presets, Proximity chat game and Discord game, set this up.
• A clearer watch folder label. It now reads "Watch folder: watch".

This release on GitHub

Fixed

• Exports no longer fail while a game is running. On a PC busy with a game and OBS, Substrike could decide it had no video encoder and refuse every export until you restarted it. It now checks again at the next export and falls back to Windows' own encoder instead of failing.
• Discord exports fit your server's limit. An export for a 10, 50 or 100 MB server is checked after it is made, and if it is still too big it is made again one step smaller, lowering the frame rate first and then the resolution, rather than squeezing the picture at full size. The outro is joined in the same pass, so it no longer pushes a clip over the limit.
• Captions stay clear of the TikTok, Reels and Shorts buttons. Vertical captions sit above the band where those apps put their buttons and text, and keep to two lines.
• Sound stays in sync. 30 fps exports could drift slightly behind the sound. Fixed.
• No captions made up from silence, music or noise. The speech model could caption a silent or music only clip with "Thank you." and the batch exported it. Those lines are now dropped when nobody is talking, and the clip's row says why it has no captions.
• A silent first audio track no longer gives a silent clip. When track 1 is silent and another track has sound, captions come from that track, the export plays it, and you are told.
• Plain error messages. A clip with no sound, a track it does not have, or a file that is not a video now says so in plain words instead of showing a technical error, with a button to open the log.
• Phone clips keep every word on screen. A rotated clip was captioned as if it were landscape, so words fell off the sides.
• Very long recordings keep their transcript. Clips longer than about 2 hours 46 minutes lost the whole transcript after transcribing.
• Smaller fixes. Vertical 1080p exports are 1080 by 1920. Opus sound now reaches the file when you pick it, and .webm exports play in browsers. A sticker no longer gets the caption box style. A backslash in your text is shown as written. Plain captions sit where the preview shows them. The Discord digest joins clips of different sizes. One badly timed word no longer stops alignment, a reel cannot be saved over one of its own clips, and the same file added twice is queued once.

Improved

• Warnings when no mic track is set. Setup, the clip summary, the editor and Batch now tell you when no audio track is marked as your mic, since that is what keeps your words apart from your friends'.
• The app asks before cancelling work. Leaving a clip while it transcribes or exports asks first instead of quietly stopping it.
• Export is always in view. The Export and Process buttons no longer scroll out of sight.
• Caption defaults that stick. Your caption choices are kept, and Batch and the watch folder use them too, including whether to add emojis.
• Settings search finds single settings, not just whole sections.
• Easier to find your way. A Help link and About sit in the header, web links open in your browser, style controls say what they do, speakers fold away once confirmed, and the tips tour waits until setup is done.

New

• Send report (optional). Settings, Privacy has a Send a report button, also offered when something goes wrong. It shows you exactly what would be sent before anything leaves your PC: the app version, Windows version, graphics card, a few settings and the end of the log with your user name, email addresses and clip names hidden. Never any video, sound, transcripts or your licence key. Nothing is sent unless you press Send.

This release on GitHub

Fixed

• Exported captions and stickers are the size you saw in the preview. Captions came out smaller in the video than in the editor, and stickers shrank too. Both now match the preview at 9:16, 16:9 and 1:1. An export with stickers but no captions no longer fails.
• Whisper's made-up stock lines are caught. Phrases like "We'll be right back." that nobody said are dropped when no one was talking there, or held for you to review when unsure.
• Montages join clips of different sizes and frame rates instead of failing.
• Exports carry one mixed audio track built from your track roles (Me, Voices, Game), with Ignore left out, so every video player plays the full mix.
• The spoken language you pick in Settings reaches batch and watch-folder clips.
• Send to Discord fits every post to your server, single clips and digests alike: 10 MB for a server with no boosts, 50 or 100 MB for boosted ones.
• Cancel, undo and restart behave. Cancelling during speaker detection stops cleanly and a failed speaker pass says so. Undoing an edit made while Align timing ran keeps the alignment. Export and Send to Discord wait while the words are being rewritten. Update and restart waits for running work, and half-finished exports are cleared after an update.
• Smaller fixes. Old highlights stay until a new scan lands, a captioned output is never picked up again as new footage, the Discord digest and held outro cards survive a restart, and the trial limit on highlight and compile exports opens the licence window.
• The crash notice's Got it button sits bottom right, where you expect it.

Improved

• Emojis are off unless you ask for them, and you can see and remove every one. Automatic emojis were on in every caption style and landed on everyday words, with no clear way to find the setting. They are now off by default, and a style you saved keeps them only if you had turned them on yourself. The preview has an Emojis on/off button, the Style tab setting is now called "Add emojis automatically", and it only reacts to strong words like lol, dead or clutch, never to names or censored words, at most one per line. In the transcript each emoji shows as a chip after its word with an x to remove it. To add your own, hover a word and click the smile button (or press E on a selected word, Ctrl+E while typing): search, or pick from suggestions, your recent picks and common gaming ones. Emojis are in full colour in the exported video too, the same pictures, size and place as in the preview.
• The update banner is one plain line saying a new version is ready, with a What's new link, instead of raw release notes.
• Align timing says plainly what it did: that all lines were re-timed and how far words moved. The automatic pass after transcribing stays quiet unless something went wrong.
• Speaker editing is quicker. Alt+S swaps the two speakers in a selection, Alt+Up and Alt+Down move words a lane, and double-clicking a word on the timeline opens it for typing.
• Substrike checks for updates every 6 hours while it is open, not only at launch.
• Lumen has his new look in the app and on the app and installer icon.

New

• 🎬 Lumen outro. Turn on "Add Lumen outro" in the export options to end a clip, montage or reel with a 5 second Substrike end card, scored with the Substrike theme. A batch digest gets it once, at the end.
• 🧭 Setup walkthrough. A short first-run guide covers your recorder, clip folder, audio tracks, your voice, language and style. Every step can be skipped, and Settings, General, Run setup again brings it back.
• 🎙️ Your mic track is found for you. For clips with several audio tracks, Substrike listens to each one and picks out your voice when it is sure, with a banner to keep that choice. When it is not sure, it asks before transcribing.

This release on GitHub

• Send to Discord fits the file to your server. A webhook can post 10 MB to a server with no boosts, so bigger clips were refused. The app now makes a smaller copy that fits and sends that. Set your server's limit in Settings, Accounts (10, 50 or 100 MB, by boost level). If Discord refuses a post, the reason stays on screen in plain words: too large, a deleted webhook, or too many posts at once. A post that overlaps another no longer loses its result.

• About shows the current third-party notices, the programs and models with their licences, instead of internal working notes from before launch. FFmpeg is now listed as the LGPL build that ships.

• Lumen keeps his outfit on the opening screen, and the opening screen fades to black before the app appears instead of showing both at once. New tagline: "Caption your clips, automatically."

• The Speakers count is respected. Setting how many people are talking with the + and - buttons now sticks, and the app says so when it finds fewer voices than you asked for.

• Voice track is picked once, in the track roles. Settings and the batch screen no longer show a separate Voice track picker next to the per-track roles. A Voice track picked before this release becomes that track's Voices role, so batches hear the same track and everyone on it is still told apart.

• Timeline word chips stay readable when zoomed out. Short words no longer turn into garbled slivers under the drag handles.

• Lumen caption style keeps spaces between words in the editor preview. Exports were already right.

• Seasonal ambience is music now, gentle chord changes instead of one long held note, and it no longer builds up notes over a long session.

This release on GitHub

First public release. Everything below from 0.3.3 is included, plus: licence keys and a 7-day trial, find-and-replace word rules, a keep-filler-words option, living captions (side by voice direction, size by loudness), rebindable keyboard shortcuts, four special themes, the Lumen caption style, SRT and ASS export, and a redesigned Settings screen.

This release on GitHub