v0.16.1

Vapourkit 0.16.1

Vapourkit now has a website along with extensive documentation!!

This release is a small follow-up to 0.16.0 — this release also bundles everything from 0.16.0 below, since the two ship back-to-back.

Bug fixes (new in 0.16.1)

• Fixed "Cannot read properties of null (reading 'execute')" crash on upscale — when something canceled or restarted an upscale while it was probing the video for frame count (manually cancelling, starting a new run from the queue, etc.), the executor reference would get nulled mid-flight and the next step would crash. The handler now captures a local reference and bails cleanly if a cancel happened during the probe. Same fix applied to the preview-segment path.
• Fixed silent stall during "Getting frame count" — on a cold BestSource cache, the frame-count probe could sit silently for a minute or more while indexing, looking like a hang. Indexing progress now streams through the same progress bar used on first video load, and gets logged to the queue item log too. No more guessing whether it's stuck.

---

Vapourkit 0.16.0

Important note:

No more manual plugin reinstall this release! Setup now installs plugins automatically as its final step, so the "reinstall your plugins from the top-left menu" dance from 0.15.0 is gone. If you're upgrading and setup detects missing plugins, it will install them on first launch — just sit through the progress bar.

33 commits since 0.15.0.

New features

• Privacy mode — new lock icon in the header that hides previews, input/output filenames, queue thumbnails and item names, and the editing-queue banner behind clickable veils. Notification toasts also become generic ("Error", "Notification", "Details hidden by privacy mode") so filenames don't leak to screen. Resolution / fps / duration stay visible. State is persisted across launches.
• Descriptive output filenames — output names now reflect your workflow instead of _processed. Example: EpisodeName-colorimetry_denoise_4x_resize2160.mkv. Includes applied filters, AI model scale, and output resolution. Auto-truncates at 32 chars. Manually choosing an output path disables auto-generation for that file. Toggle lives in Settings → Processing. Thanks @fs10102020!
• No-filters safety — a persistent banner now sits above the Upscale button when no filters are enabled, describing exactly what encoding will do, and a confirm dialog fires if you try to upscale anyway. The old "default-upscale" silent fallback (which would secretly run whichever AI model happened to be selected first) has been removed.
• Video indexing progress bar — BestSource indexing progress now streams to the UI as a progress bar under the video drop zone, so you can see what's happening on long initial loads instead of staring at a frozen panel.

Setup & installation reliability

• Auto-install plugins at end of setup — plugins are installed as the last step of first-run setup, with auto-retry-once on transient failures and explicit Retry / Continue-without-plugins recovery if it hard-fails. The PluginsModal stays for ongoing management from the menu.
• Fixed "Failed to initialize VSScript" on fresh installs — pinned vapoursynth==72 and vsjetpack==1.1.0 to prevent pip from silently upgrading the bundled Python binding to a newer ABI-incompatible version. Recent vsjetpack releases require vapoursynth>=73, which was clashing with the bundled R72 runtime DLL.
• Fixed vsview launch error — switched from python -m vsview to the vsview.exe Scripts wrapper (the vsview package has no __main__.py).

Filters

• Temporal Fix filter renames — Temporal Fix V2 is now TemporalFix (AI), and the classic version is now TemporalFix (Classic). Filter definitions revised.

UI & UX

• Descriptive-naming regen now respects your configured default output folder (was previously clobbering it with the input video's directory on first render).
• No-filters confirm dialog replaced with a persistent banner — fewer dialog interruptions, clearer wording about what encoding will actually do.

Under the hood

• New BestSource stderr progress parser with tests covering lastIndex reset and negative-percentage edge cases.
• New video-index-progress IPC channel; stdout/stderr handlers split in the frame-count probe to preserve descriptive error messages.
• pluginInstaller progress is routed through setup-progress while in setup mode, with a new retry-setup-plugins IPC and installing SetupProgress variant.
• New installDependenciesForSetup with auto-retry-once; "All Dependencies complete" emission moved out of dependencyManager.
• Content-length header narrowed to string before parseInt (fixes type error under newer @types/axios).

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

Consider joining the Discord server! Alternatively, we now have a Fermi/Spacebar guild.

v0.16.0

Vapourkit 0.16.0

Important note:

No more manual plugin reinstall this release! Setup now installs plugins automatically as its final step, so the "reinstall your plugins from the top-left menu" dance from 0.15.0 is gone. If you're upgrading and setup detects missing plugins, it will install them on first launch — just sit through the progress bar.

33 commits since 0.15.0.

New features

• Privacy mode — new lock icon in the header that hides previews, input/output filenames, queue thumbnails and item names, and the editing-queue banner behind clickable veils. Notification toasts also become generic ("Error", "Notification", "Details hidden by privacy mode") so filenames don't leak to screen. Resolution / fps / duration stay visible. State is persisted across launches.
• Descriptive output filenames — output names now reflect your workflow instead of _processed. Example: EpisodeName-colorimetry_denoise_4x_resize2160.mkv. Includes applied filters, AI model scale, and output resolution. Auto-truncates at 32 chars. Manually choosing an output path disables auto-generation for that file. Toggle lives in Settings → Processing. Thanks @fs10102020!
• No-filters safety — a persistent banner now sits above the Upscale button when no filters are enabled, describing exactly what encoding will do, and a confirm dialog fires if you try to upscale anyway. The old "default-upscale" silent fallback (which would secretly run whichever AI model happened to be selected first) has been removed.
• Video indexing progress bar — BestSource indexing progress now streams to the UI as a progress bar under the video drop zone, so you can see what's happening on long initial loads instead of staring at a frozen panel.

Setup & installation reliability

• Auto-install plugins at end of setup — plugins are installed as the last step of first-run setup, with auto-retry-once on transient failures and explicit Retry / Continue-without-plugins recovery if it hard-fails. The PluginsModal stays for ongoing management from the menu.
• Fixed "Failed to initialize VSScript" on fresh installs — pinned vapoursynth==72 and vsjetpack==1.1.0 to prevent pip from silently upgrading the bundled Python binding to a newer ABI-incompatible version. Recent vsjetpack releases require vapoursynth>=73, which was clashing with the bundled R72 runtime DLL.
• Fixed vsview launch error — switched from python -m vsview to the vsview.exe Scripts wrapper (the vsview package has no __main__.py).

Filters

• Temporal Fix filter renames — Temporal Fix V2 is now TemporalFix (AI), and the classic version is now TemporalFix (Classic). Filter definitions revised. Thank you @pifroggi for TFV2 and your guidance in implementing it!

UI & UX

• Descriptive-naming regen now respects your configured default output folder (was previously clobbering it with the input video's directory on first render).
• No-filters confirm dialog replaced with a persistent banner — fewer dialog interruptions, clearer wording about what encoding will actually do.

Under the hood

• New BestSource stderr progress parser with tests covering lastIndex reset and negative-percentage edge cases.
• New video-index-progress IPC channel; stdout/stderr handlers split in the frame-count probe to preserve descriptive error messages.
• pluginInstaller progress is routed through setup-progress while in setup mode, with a new retry-setup-plugins IPC and installing SetupProgress variant.
• New installDependenciesForSetup with auto-retry-once; "All Dependencies complete" emission moved out of dependencyManager.
• Content-length header narrowed to string before parseInt (fixes type error under newer @types/axios).

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

Consider joining the Discord server! Alternatively, we now have a Fermi/Spacebar guild.

v0.15.0

Vapourkit 0.15.0

Important note:

Reinstall your plugins in the top left menu when you install this version!!! If you don't do this, you WILL run into issues and you will be missing most of the new filters available in this release!

26 commits since 0.14.0.

New features

• Temporal Fix V2 — AI model based temporal coherence filter with CPU, CUDA, and TensorRT backends. Faster and more accurate than the classic mvtools-based version. Both sit side-by-side so existing workflows keep working. Thanks @pifroggi!
• Benchmark mode — measure processing throughput without committing to a real output.
• GPU monitoring — live GPU usage surfaced in the UI.
• ETA + taskbar progress — estimated time remaining visible in-app, and progress mirrored on the Windows taskbar so you can track jobs without keeping the window focused.
• Skip validation button — bypass setup validation when you already know the environment is good.
• Auto-detect now supports all model types — model auto-detection works across the full supported model list, not just a subset.

Model selector rework

• Model manager and model selector merged into a single unified modal.
• Import modal reworked and more forgiving.
• Default model categories updated.
• Auto config updates applied on first launch of a new version.

UI & UX

• Recents list reliability fixes.
• Segment selector now reports the correct frame count.
• Video panel bug fixed.
• Multiple modal polish passes — fewer rough edges, fewer stuck states.
• "Unbuilt text" rendering fix.

Under the hood

• Major App.tsx refactor — extracted action buttons, modal rendering, progress panel, and app-level effects into dedicated components and hooks. Queue state split cleanly into store / operations / effects layers.
• Test infrastructure added.
• Fixed bug with model extraction.
• Fixed bug causing unnecessary setup re-runs.

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

Consider joining the Discord server! Alternatively, we now have a Fermi/Spacebar guild.

v0.14.0

Hi everyone! Another major release! A lot changed during development between these releases, so we've skipped version or two.

This release is dedicated to my Mom. She passed away on 1/1/26 after a long battle with small cell lung cancer. Rest in peace 💖.

Important note:

Reinstall your plugins in the top left menu when you install this version!!! If you don't do this, you WILL run into issues and you will be missing most of the new filters available in this release!

0.14.0

• Adds over 150 new filters, including many from Hybrid!
• Adds a new model, AniRestore. Ideal for restoring older anime or cartoons with dot crawl, rainbows, noise, or blur!
• Replaces the filter selection dropdown with a new modal
  • This has a tag system to make finding filters easier
  • It also has a search!
• Fixes the lag and focus issues present in previous versions of Vapourkit
  • You can now have 20+ filters expanded in your workflow and it will not slow down!
  • The bug that required alt tabbing to fix is no longer present
• Adds vs-view! This allows for realtime previewing of how your video will turn out without having to render the whole thing
• Adds ESC button support to all pop up modals
• Added vs_grain
• Replaces pop up dialogs with notifications within the GUI
• Backend change: change legacy "TSPAN" text to "VSR". This change was made in conjunction with releasing TFDAT, which effectively replaces TSPAN + TSPANv2
• Lots of GUI tweaks and bug fixes to make it more cohesive and consistent
• Add option to Settings to set a permanent output path for all videos
• Add option to duplicate queued items, and overhaul the behavior of the queue button
• Added some scripts for the building process to make deploying releases easier
• Fixed vkworkflow exporting not saving encoding settings, and also removed weirdly specific settings from the templates
  • Also added support for the new category/tag system for filters

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

Consider joining the Discord server! Alternatively, we now have a Fermi/Spacebar guild.

v0.12.2

Happy holidays to everyone!! This release ships with a new upscaling model intended for older SD quality anime. It's capable of fixing quite a lot of issues: https://slow.pics/c/lzVKwDRv

0.12.2

• Fix BF16 engine names (previously appended _fp16 when it's _bf16)
• Remove unused code
• Hide Validate button during processing
• Rename Color Matrix to Colorimetry as it does more than the name implies
• Improve GUI responsiveness
• Change the way Developer Log works. It now polls main.log instead of printing directly to the UI
  • This also has the added benefit of fixing formatting issues that were present previously
• Include 2x_bndl_animefilm_v1.5 FDAT

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

v0.12.1

This release contains an important bug fix. If you were running into errors running any workflows, please update immediately.

Happy holidays!!

0.12.1

• Overhaul validation method. It will no longer automatically run in the background, instead you must manually run it if desired
  • This fixes a lot of the lag present in older versions of Vapourkit
• Fix issue where "Same as Input" was the default for fresh installs of Vapourkit
• Fix broken AV1 presets

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

v0.12

This release contains TONS of improvements, fixes, and QoL changes. This is a direct jump from v0.10.1 to v0.12, there were no releases for versions in-between. Thanks so much to everyone who tested the development builds on Discord, and especially @pifroggi for helping a lot with the filters and fixing color issues.

Happy holidays to everyone!!

Highlights:

• Overhauled the Batch Processing panel, now displays thumbnails for each clip
• RIFE support
• Segment support - Cut a video into a smaller segment to process
• Preview your workflow on a clip! The preview feature is contained in the segment selection area
• Switch to vs-mlrt 15.13, which provides a small speed bump
• Fix lots of issues that affected general performance and usability
• Overhaul TensorRT engine building - Now has BF16 support, automatic shape detection for static models, and more!
• Remove Simple mode. Apologies to anyone who used it, but it was complicating the code base quite a lot
• Add an easy panel to control ffmpeg settings!
• Reduce issues with color shifting on outputs

Changelog:

0.12.0

• Remove simple mode to simplify codebase
• Move encoding settings from Settings panel to the right pane, and add easy toggles for common settings
• Add RIFE filter for frame interpolation
• Overhaul vkfilter parsing to be more robust
• Fix GUI design inconsistencies
• Reverted to vs-mlrt 15.13 as 15.14 has noticeably lower performance
• Update vs_tiletools
• Update zsmooth to 0.15
• Add an easy panel to control ffmpeg settings!
• Reduce issues with color shifting on outputs

0.11.0

• Change the way file names are handled for models
• Overhaul the design of the header to save space
• Move the DirectML toggle from Settings to the header
• Change the default model type from tspan to image to reduce chance of error for models without metadata
• Fix audio clipping when using segments
• Allow users to customize video-compare settings in the Settings menu
• Force kill trtexec and vspipe processes when beginning workflow processing
• Add MC_Degrain filters
• Change to vs-mlrt version 15.14 from 15.13 RTX
• Add detection for vs-mlrt version changing (will not take effect in this release)
• Add BF16 toggle when building TensorRT models
• Add automatic static + shape detection when building TensorRT models
• Add update system for vs-mlrt plugin
• Update vs_undistort to version 2.0.0 (thanks tepete!)
• Update queue panel behavior and design to be more intuitive

0.10.2

• Implement segment selection. Users can now select a small segment of a video to process and preview!
  • When using this mode, the comparison buttons are disabled
• Fix issue where highlighted code wasn't visible in the Filter panel
• Minor bug fixes
• Fix GUI lag
• Add search function to Manage Models menu

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

v0.10.1

This release implements batch processing, queue management, update checking, and contains a bunch of other QOL improvements!

0.10.1

• Redesign "Show Queue" button and change location
• Allow the user to change the color space the output video is saved in
• Rework the Settings menu to be easier to use
• Fix the way videos are displayed when processing is complete

0.10.0

• Add batch video processing support
• Add ability to launch comparisons in from queue list
• Add experimental update checker
• Add force stop button for stuck processes
• Clean up About menu
• Improve changelog display
• Fix processing bug with batch processing
• Update zsmooth plugin to 0.14
• Overhaul internal code for start/stop processing button
• Overhaul Video Info Panel
• Add documentation for Batch Processing
• Shrink queue panel and clean up unused files
• Fix color scheme of syntax highlighting

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

v0.9.4

Changelog:

0.9.4

• Clarify precision options in GUI
• Add syntax highlighting for filters
• Add section for license information of included models
• Add link to GitHub page in About window

---

To help fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download Vapourkit

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!

v0.9.3

To encourage donations to fund development, I have chosen to host the binaries for free on my Ko-fi page here:

Download

Vapourkit is 100% free, and there are build instructions included in the GitHub repository if you'd rather build it yourself!