Changelog

v0.13.0 — 2026-07-26

• Far fewer visible jump cuts. When Smart Switch removes a silence and cannot hide the join by cutting to another angle, it now extends the camera you are already on straight across the join — but only when the person on screen is not speaking, so lips never fall out of step with the audio. It also refuses the trick whenever it would show more than a couple of seconds of stale footage. On our 71-minute, 3-speaker reference episode, visible jump cuts dropped from 27 to 6.
• New setting: “Maximum time on a listener”, under Smart Switch, set to 5 seconds by default. Cutting to someone reacting is normal and stays; what this stops is the edit parking there. Holding one camera across a run of nearby cuts could previously leave you watching a silent face for up to 12 seconds — now the edit returns to whoever is talking once your limit is reached. Worst case on the reference episode went from 12 seconds to 5.
• Those two improvements did not come at each other's expense. Shortening the long listener shots removed three more jump cuts rather than adding any back.
• Behind the scenes: the regression suite that replays real episodes on every change (16 of them, 2 to 14 cameras, 2 to 72 minutes) now tracks time spent on a listener alongside jump cuts, and checks both together — so neither can quietly be improved at the other's cost.

v0.12.0 — 2026-07-23

• Far fewer visible jump cuts after silence removal. Smart Switch now proves every removed-silence seam is hidden before it ships: it cuts to a different angle at the seam, invisibly extends the other camera across it, or places a short reaction ahead of it. Long pauses, back-to-back pauses, and speaker handoffs with a trailing “yeah, exactly” now resolve cleanly — handoffs land as a natural L-cut, with the next speaker already on screen as they start talking. On our 40-minute reference episode, visible jump cuts dropped from 25 to 5.
• The camera no longer lingers on someone listening while the other person talks. Camera returns now aim at whoever is actually speaking over the next few seconds, and a final pass catches any long stuck shot the edit missed — the worst case in testing went from 18 seconds on the wrong person to 7.
• Fewer rapid back-and-forth camera flips around dense clusters of cuts.
• Behind the scenes: every one of these improvements is locked in by a regression suite that replays 13 real episodes (2–14 cameras, 8–72 minutes) on every change, so edit quality can only move forward.

v0.11.1 — 2026-07-20

• Fixed a camera-switching bug where the edit could open on the wrong person. When you removed silence near the start of a section, the camera used to stay on whoever spoke last instead of cutting to who is actually talking — now it follows the active speaker from the very first cut. Multi-speaker interviews and podcasts that trim dead air are the main beneficiaries.
• Same-speaker pauses still merge cleanly, so this does not add extra cuts when one person is talking through short gaps.

v0.11.0 — 2026-07-17

• The panel now checks your timeline before every run. If nothing is mapped to a speaker mic, Smart Switch has no camera to work with, or your mapped mic tracks turn out to be empty, it stops and tells you exactly what to fix — in plain English — instead of spending minutes producing a bad edit.
• The check is instant, re-reads your sequence at the moment you click Run (so edits you just made in Premiere count), and stays completely silent when your setup is fine.
• Runs now cancel cleanly if you switch to a different sequence while one is starting.

v0.10.0 — 2026-07-17

• A redesigned panel. A new home screen puts Quick Run one click away, the Full Pipeline is now a guided wizard, and every tool asks fewer questions before it runs.
• Presets: save your show's speaker & camera mapping and your tool settings once, then reapply them on next week's episode with one click.
• Smart track detection: the panel now reads your sequence — including nested multicam setups — and fills in speakers, cameras and routing automatically. Everything it detects is marked AUTO; click any speaker name to rename it, and the camera routing follows the new name.
• It also recognises when a sequence is a previous Multicam Toolbox output and suggests running on the original instead.
• The panel now follows Premiere: switch to a different sequence and it returns home with fresh details instead of showing the old sequence's tracks — and tools will not run against a different sequence than the one they were set up on.
• A short first-run tour and an in-panel help drawer, so you do not need the manual to get to your first cut.

v0.9.0 — 2026-07-15

• Repairs automatic updates so future fixes reach you on their own — this is the release that makes the self-update system actually work end to end.
• Adds a reminder for anyone still on an older version to update via the Adobe Creative Cloud app.
• Behind the scenes: more reliable crash and usage reporting so problems get found and fixed faster.

v0.8.12 — 2026-07-14

• Fixes automatic updates. Until now the plugin could download an update, report success, and then keep running the old version anyway — so fixes shipped in recent releases never actually reached the panel. This is the version that repairs the updater itself, so future fixes can arrive automatically.
• Includes the panel-goes-dark and idle-CPU fixes from the last two releases, now that they can actually be delivered.

v0.8.11 — 2026-07-14

• Fixes the panel going dark. Clicking a checkbox — most often "Trim welded filler words" in the Full Pipeline, but also the track checkboxes in Audio Export, Silence Removal and Filler Removal — could scroll the whole panel out of view, leaving it looking blank and dead until you reloaded it. Your work was never lost; the panel just wasn't where you could see it. All of those checkboxes are safe to click now.
• The panel no longer burns a CPU core while it sits idle, so Premiere should feel snappier whenever it's open.

v0.8.10 — 2026-07-14

• Fixes a problem introduced in 0.8.9 where camera switching could leave one-frame flashes on the timeline, and could cut up or switch off clips on your title, overlay and b-roll tracks. Only the camera tracks you map are touched now — everything else on your timeline is left completely alone.
• The "Maximum camera time" limit is now applied exactly, instead of occasionally overshooting.

v0.8.9 — 2026-07-13

• Camera switching no longer leaves the wrong person on screen. In some spots the edit could sit on the listener for 30 seconds or more while the other host was talking — that is fixed, and short reaction shots are untouched.
• Your "Maximum camera time" setting now actually applies. Shots could previously run for well over a minute on a single angle during a long answer; the edit now breaks them up with a wide shot and comes back. The cut is placed on a natural pause rather than exactly on the limit, so it does not feel mechanical.
• Windows: filler-word removal is noticeably more accurate. The speech recognizer was falling back to coarse timing on short clips, which caused it to miss fillers it should have caught.

v0.8.8 — 2026-07-13

• Filler-word removal now works out of the box. The speech recognizer it depends on was missing from the installer, so the tool could fail with a "whisper-cli was not found" error unless you happened to have it installed yourself. It is now set up automatically the first time you use the tool — a small one-time download.
• Filler detection is more accurate: each candidate is now checked individually against the speech recognizer, so fewer real words get cut and more genuine "um"s get caught.
• New option to tidy up fillers that are welded onto the neighbouring word, with a short fade at the cut so the edit does not click.
• Camera switching: better choices at existing cuts, fewer awkward returns to the previous camera, and a fix for short same-speaker segments being skipped.

v0.8.7 — 2026-07-07

• Voice mixing sounds much more consistent: loud, excited moments keep their energy instead of getting pushed down, quiet asides come up naturally, and the overall level no longer drifts around during an episode. Moments where hosts talk over each other also keep their natural punch.
• The per-track Compression setting in voice processing now actually applies (it was silently ignored before) and has been retuned to smooth out sharp syllable peaks without squashing the performance.
• Filler-word removal no longer clips fragments out of longer words.
• Camera switching: fixed rare cases where healing cuts could merge wrongly around short gaps, and where planned cuts were only partially applied.

v0.8.6 — 2026-06-16

• Silence removal: you can now pick extra audio tracks (like a soundboard or sound-effects track) to count as sound, so they are no longer mistaken for dead air and cut. Your speaker tracks are always included by default.
• Fixed a bug where the automatic safety backup of your sequence could silently fail to appear before edits ran. Backups are reliable again, and the Voice Mixing tool now makes one too.
• Restored anonymous, opt-in usage insights that had stopped sending. No change to what is collected or your privacy settings.

v0.8.5 — 2026-06-13

• Continued behind-the-scenes improvements to anonymous, opt-in usage insights so we can understand which tools get used and where they fall short. No change to what's collected or your privacy settings.

v0.8.4 — 2026-06-13

• Behind-the-scenes improvements to anonymous, opt-in usage insights so we can see which features matter most and prioritize what to build next. No change to what's collected or your privacy settings.

v0.8.3 — 2026-06-13

• The Full Pipeline's voice processing now uses the same gentle, natural-sounding noise reduction as the standalone Voice Mixing tool — no more hollow or "phone-like" quality on automated runs.
• You can still choose stronger denoising explicitly when a track needs it.

v0.8.2 — 2026-06-13

• Behind-the-scenes reliability fixes to the build and auto-update pipeline so future improvements reach you smoothly.
• Restored the anonymous diagnostics that help us spot and fix issues faster (opt-in, unchanged from before).

v0.8.1 — 2026-06-13

• Behind-the-scenes reliability fixes to the build and auto-update pipeline so future improvements reach you smoothly.
• Restored the anonymous diagnostics that help us spot and fix issues faster (opt-in, unchanged from before).

v0.8.0 — 2026-06-13

• Voice processing now sounds noticeably more natural and closer to professionally produced audio.
• Reworked noise reduction removes the muffled, "phone-like" quality from earlier versions for clearer, fuller-sounding speech.
• The loudest moments stay clean instead of sounding harsh or crushed, keeping each voice's natural dynamics.
• Overall loudness is better matched to professional podcast delivery levels.

v0.7.7 — 2026-05-14

• Filler detection is more accurate: candidates are scored in conversational context with an SNR sanity check to reject false positives in quiet regions.
• Short filler removals now ripple-delete with cut-boundary fades for cleaner edits.
• The voice tool gets a higher-quality output pipeline with true-peak limiting, 32-bit float processing, and stereo upmix.
• The filler widget now shows real candidate data from your run.
• Sequence backups are created once before any pipeline edits, so re-runs share a single safety snapshot.

v0.7.5 — 2026-04-28

• The ToolRunOverlay now has a unified design, showing clear states for running, completed, and errors, making it easier to manage tools.
• You can quickly dismiss or return to the home screen from the ToolRunOverlay for better navigation.
• The app version is now displayed more reliably, ensuring you always know which version you're using while editing.

v0.7.4 — 2026-04-27

• Enjoy a more consistent interface with the new unified ToolRunOverlay, featuring clear indicators for running, completed, and error states across all tools.
• Easily dismiss or return to the home screen from any tool with the updated overlay controls.
• Benefit from improved performance, as the version number is now resolved at build time for a smoother experience.

v0.7.3 — 2026-04-27

• The ToolRunOverlay has been updated for a more cohesive look and feel, now showing running, done, and error states consistently across all tools.
• You can easily manage and dismiss actions from the ToolRunOverlay, enhancing user navigation.
• Version information is now more reliably displayed, ensuring you’re always aware of the current version during your editing sessions.

v0.7.2 — 2026-04-25

• Unified ToolRunOverlay now clearly shows running, done, or error states, making it easier to track the progress of all tools.
• Dismiss and home actions are now available across all tools for improved usability.
• The app version is now displayed correctly at build time, ensuring you always know which version you're using.

v0.7.1 — 2026-04-24

• The ToolRunOverlay has been updated for a more consistent user experience, showing clear states for running, completed, or errors across all tools.
• You will now see the app version resolved at build time, enhancing performance and reliability during use.

v0.7.0 — 2026-04-23

• Plugin now self-updates in the background — smaller fixes will apply automatically without reinstalling the ZXP.
• Settings → About now shows both the installed plugin version and the currently-running bundle version for easier support.

v0.6.0 — 2026-04-22

• The ToolRunOverlay has been unified, providing clear states for running, completed, and error actions across all tools.
• You can now easily dismiss the overlay or return home, streamlining your workflow.
• The app version is now determined during the build process, ensuring consistent versioning throughout your editing session.

v0.5.0 — 2026-04-22

• A unified overlay now shows clear states for running, completed, and error messages for all tools, along with options to dismiss or return home.
• The plugin version is now displayed at build time for a more reliable experience.

v0.4.6 — 2026-04-22

• The ToolRunOverlay now presents a unified interface with clear indicators for running, completed, and error states, enhancing usability across all tools.
• You can easily dismiss or return to the home screen from any tool with streamlined actions.
• The app version is now resolved at build time for more accurate display and improved performance.

v0.4.5 — 2026-04-21

• Automated changelog pipeline — release notes are now generated automatically on each release and summarized in plain English
• New What's New screen in the panel shows release highlights once per version

v0.4.4 — 2026-04-21

• New unified status overlay across all tools — shows progress while running, then a done or error screen with options to dismiss or go back to the home screen
• Version number is now resolved at build time for more reliable display

v0.4.3 — 2026-04-18

• Replaced preset buttons in Voice Mixing with direct controls for noise reduction, compression, and EQ — more precise tuning without guesswork
• Smoother voice dynamics — the compressor now holds before releasing, preventing pumping on short pauses between words
• More natural speech-adaptive processing that responds to your specific recording environment
• Reduced aggressiveness in per-track leveling to preserve natural volume variation

v0.4.2 — 2026-04-14

• Voice Mixing is now clearly marked BETA and disabled by default — opt in when you're ready to experiment

v0.4.1 — 2026-04-12

• Fixed signing compatibility issues with the ZXP installer on some systems
• Fixed macOS quarantine prompt on first launch

v0.4.0 — 2026-04-12

• New panel design with individual tool screens — Smart Switch, Silence Removal, Voice Mixing, and Audio Export each have their own dedicated view
• Full Pipeline wizard: run all three processing tools back-to-back in a single pass
• Help button in each tool links directly to the relevant section of the guide

v0.3.0 — 2026-04-10

• Improved voice leveling pipeline with adaptive dynamics
• In-panel Help button added — opens the user guide to the relevant section

v0.2.x

• Smart Switch, Silence Removal, and Voice Mixing as standalone tools
• Audio Track Export for individual WAV stem extraction
• License gate with key validation