Help us keep the list up to date and submit new video software here.
Tool
Complete Version history / Release notes / Changelog / What's New for VidControl
v2.9.31
2026-06
Redesigned toolbar: tidy groups instead of a cramped icon strip, rare tools tucked into a "" menu, and watch-folder actions now sit next to the watch toggle — readable in both dark and light mode
Reworked track selection rules: clear groups (Audio / Subtitles / Burn-in), plain language instead of cryptic terms, a single dropdown for forced subtitles — and the "forced subs matching the audio language" mapping now actually works (e.g. Japanese audio German forced subtitles for signs/songs)
Per-preset file naming: each profile remembers its own naming scheme, so a "Series" profile names episodes differently from a "Movies" profile — including for files auto-encoded from watch folders
Settings that weren't being remembered (notifications, sound, taskbar flash, conflict action and more) now persist; the "parallel jobs" slider works again
The Command tab now shows the real, runnable ffmpeg command (including 2-pass) and refreshes automatically when opened
Surfaced hidden features: VAD/pacing fine-tuning and the diagnostics export now live in the Settings tab
Smaller fixes: only one "Support VidControl" prompt after batch encodes, a de-cluttered Queue tab, filled-in translation gaps, and presets no longer lose their track rules when saved
v2.9.30
2026-06
Watch folders: a new switch, Auto-encode detected files. With it on, anything dropped into a watched folder is encoded automatically — previously you had to start each file by hand. Off keeps the old behaviour: files just go into the queue and you start them yourself
Removed the Encoding report window that popped up after each batch. It blocked the start button until you clicked it away. The Windows notification stays, and the full summary is still in the Log tab
v2.9.29
2026-06
New: AI dubbing. VidControl can now voice your translated subtitles itself. It pulls the voices out of the video, clones one per speaker and builds a complete new audio track. The music and sound effects stay intact, and the finished track is added to the video as an extra audio track on your next encode
New: Dub Studio for hand-tuning. Every voice sits on its own lane on a timeline. You can listen to individual lines, move, shorten, raise or lower them and fade them in and out. If lines run into each other, one click on 'Resolve overlaps' pulls them apart. Single lines can also be regenerated
New: audition a voice first. Before the long run starts, pick a speaker, type any sentence and hear right away whether the cloned voice works
AI dubbing is optional and runs entirely locally on your graphics card. VidControl downloads what it needs by itself the first time
Clearer codec list: codecs that can't run in the current mode (e.g. VVC or AV1 under NVENC) now show greyed out with a reason instead of just being missing, so you can see why something isn't selectable
Fixed the dub-engine setup: the window now closes by itself after the download and continues straight on; the close button no longer hangs
v2.9.28
2026-06
Pulled before release due to a dub-engine setup soft-lock — all of its features ship in 2.9.29
v2.9.23
2026-05
NEW — Hybrid transcription: choose the timing model (segmentation & timestamps) and the text model (wording) separately. Perfect when a general model like large-v3 times speech well but a language-specialised model like Belle understands the language better — it gives you the good timing AND the good words. Directly fixes the case where Belle assigns single words multi-second cues
The two models run one after another, so even a single GPU under 16 GB only ever loads one at a time. A recommendation is shown based on your graphics card (VRAM) — but every choice is yours; nothing is hardcoded
The text model's words are aligned onto the timing model's segments by content (character alignment), not by the text model's untrusted timestamps
v2.9.22
2026-05
Fixed the readability re-segmentation so it actually fixes heavily fragmented subtitles (one word per cue — common with the Chinese model + word-by-word translation). A reading-speed gate had been blocking the very merges it needed; word-cues are now correctly merged back into full sentences
Added a timing step that extends merged cues into the following gap until they are comfortably readable (reading-speed target + minimum duration), never overlapping the next cue
Splitting now triggers only when a cue is on screen too long or has too much text to fit — not on reading speed, which more cuts can never improve
v2.9.21
2026-05
NEW — Context-aware translation: the LLM now sees the already-translated neighbouring cues (a rolling window) plus a few upcoming source lines, instead of translating each line in isolation. This is the biggest readability win for translated subtitles — terminology, names and register stay consistent across chunk boundaries, and sentences that span several cues read naturally
On by default (4 context lines); tunable via the TranslationContextSize setting (0 turns it off). Context lines are clearly marked "do not translate", so the index mapping and output are unchanged
v2.9.20
2026-05
NEW — Readability re-segmentation: a decoupled engine that re-cuts subtitles so the text reads naturally — merges fragmented cues, splits over-long/too-fast ones at the most natural boundary (sentence comma conjunction word), and rebalances lines, with separate CJK and Latin rules and proportional time redistribution
Runs automatically after translation (on by default, with a toggle in the translation dialog) — fixes the common case where translated German breaks at the original Chinese cue boundaries instead of natural German ones
Also available as a "Re-segment" button in the subtitle editor (Fix tab), reversible with Ctrl+Z; works on any subtitles — imported SRT, OCR output, external translations, batch repair
Translation dialog is now scrollable with the buttons pinned at the bottom, so the "Translate" button is always visible even on small or scaled displays
v2.9.19
2026-05
NEW — Target-VMAF auto-CRF: VidControl sample-encodes a few representative segments and binary-searches the highest CRF (smallest file) that still reaches your quality target (e.g. VMAF 95). One button in Advanced Video settings opens a live probe dialog and applies the chosen CRF to your preset
VMAF analysis fixed: an unescaped Windows drive-letter colon in the libvmaf filtergraph meant VMAF never actually ran; also made resolution-robust via scale2ref
Fixed a startup crash in 2.9.18: duplicate localization keys (left over from the i18n waves) threw before logging initialized, so the app exited with no error dialog and no log
Diagnostics export now redacts webhook URLs (not just keys/tokens)
Subtitle editor: pacing now operates on the correct file if you switch tracks during the audio analysis; removed double event-subscriptions on track load; QC auto-fixes pick the correct neighbour cue when two cues share a start time
Settings are saved thread-safely — structural list mutations during parallel encodes can no longer make a debounced save silently fail
v2.9.18
2026-05
Command palette (Ctrl+Shift+P): fuzzy-search every command, preset and recent file — any of the ~119 actions reachable in two keystrokes
Universal undo/redo in the subtitle editor (Ctrl+Z / Ctrl+Y) across all ~28 mutating operations — batch and AI edits are now safely reversible
Subtitle QC linter with one-click "fix all" (Netflix/EBU profiles, dedicated CJK reading-speed & line-length limits): reading speed, line length/count, min/max duration, overlaps and gaps
Snap-to-silence activated: cue edges now align to real detected audio gaps in both the editor and the Whisper flow; waveform peaks are cached to disk so reopening is instant
Crash-safe settings: atomic writes (no more truncated settings.json), debounced auto-save, and a redacted diagnostics .zip export for bug reports
Encoding: HDR10 static metadata (MaxCLL + mastering-display) is now preserved on libx265 output, AV1 film-grain synthesis, an encoder -tune selector, sharper Lanczos downscaling, +faststart for MP4, NVENC spatial/temporal AQ, and a VMAF analysis fix that made it work on Windows at all
Translation: pinned term overrides ("this name always becomes that") are now enforced after the LLM, a free-text style/register instructions field, and a higher Whisper decode-quality floor
v2.9.17
2026-05
FFmpeg setup dialog on first run: auto-download from BtbN (~80 MB), point at existing ffmpeg.exe, or skip for now — no more "where do I get FFmpeg" wall for new users
Extended FFmpeg detection chain: settings path tools/ffmpeg/ tools/ffmpeg/bin/ PATH common install locations
FFmpegProcessManager hot-swappable path — after auto-download, encoding becomes available without an app restart
CLAUDE.md updated: FFmpeg is deliberately not bundled to keep VidControl decoupled from FFmpeg's monthly release cycle and avoid LGPL/GPL attribution overhead
v2.9.16
2026-05
Video codec dropdown now filters by hardware capabilities — codecs your NVENC/AMF/QSV can't actually encode (e.g. VP9 on any GPU, AV1 without Ada/RDNA3/Arc) are no longer selectable, preventing silent CPU fallbacks
ASS/SSA subtitle burn: embedded font/size/outline from the source file is now honored by default instead of being overridden with Arial (new override checkbox if you want the old behavior)
HEVC 10-bit on Pascal NVENC: detects the GPU can't encode 10-bit HEVC and converts to NV12 automatically instead of silently dropping to 8-bit
Settings loader now retries on file locks (OneDrive sync, antivirus, second VidControl instance) instead of falling back to defaults and silently losing user settings
Subtitle editor distinguishes save errors: read-only file vs. locked by another process vs. generic — clearer feedback
Whisper process kill uses a retry loop on cancel — no more zombie processes
UTF-16 subtitle files without BOM are now detected (previously garbled text)
Custom FFmpeg audio args now sanitized the same way video args are
FilePath null validation in the encoding pipeline — explicit error instead of cryptic "no such file" from FFmpeg
Resegmenter: crash guard for files with 1 entry
v2.9.15
2026-05
Complete English localization — every dialog, status message, queue state, and error text fully translated (around 400 strings localized across two waves)
OCR default language now follows the UI language (English UI English OCR; previously always German regardless of UI)
First-run language derived from your OS culture instead of forcing English
Output folder selector promoted to the main toolbar with browse + one-click „use source folder" reset (was previously buried in Settings)
Whisper VAD parameters (threshold, min silence, max speech per segment) exposed in Settings for power users
Belle Chinese model made usable — Silero VAD pre-segmentation forces sub-cues 12s instead of 30-60s blocks
CJK-aware post-splitting: Chinese subtitles split at with character-proportional time distribution
Whisper hallucination detection: phonetic-floor heuristic auto-drops garbage cues under 0.40s
Pacing review dialog with two tabs (hallucinations + merges) and audio preview per cue
Update system fixed — was reporting 2.7.4 as latest despite newer 2.9.x releases
Ctrl+H opens Find & Replace in the subtitle editor; tooltips added for video player and sync point actions
v2.7.4
2026-03
EAC3/Atmos audio in M4V output: FFmpeg muxer fix prevents exit code -22
Intro detection retroactive: Already loaded files are scanned when enabling the feature
Fixed pipe deadlocks in 5 services (DoviTool, AutoSync, PostProcessing etc.)
Crash safety: fixed async-void crash in GPU monitor and update dialog
Thread safety & resource leak fixes across 22 files
OCR spell correction: Replacements now context-aware per word
PropertyChanged notifications for file size, bitrate, speed, ETA
v2.7.3
2026-02
Remove scene tags: Option to automatically strip release tags from filenames (e.g. German AC3 480p WEBRip XviD-FND)
Queue context menu: Open output folder, open source file, move output file
PDB files are no longer generated in release builds
v2.7.2
2026-02
Custom titlebar: Toolbar is now the title bar — saves one row of height
Min/Max/Close buttons clearly visible in purple toolbar (dark theme fix)
Status bar at bottom doubles as drag area for window movement
v2.7.1
2026-02
Minor bugfixes and stability improvements
v2.7.0
2026-02
Settings decluttered: Progressive disclosure with collapsible expanders
Advanced Video, Custom Filters, Expert Options, Post-Processing now collapsed by default
Track Selection Rules: Basic options visible, advanced options in sub-expander
~40% less visual noise in Settings — no features removed, just hidden for power users
Fixed: Auto-update check did not detect new versions due to HTTP caching — please update manually this one time
v2.6.5
2026-02
Whisper Review: Spell check & consistency review after transcription
Interactive spell check with Hunspell (same as OCR review)
Consistency checks: duplicate entries, high CPS, short duration, inconsistent spellings
Automatic language mapping (Whisper Hunspell dictionaries)
v2.6.2
2026-02
Whisper dialog: All inputs locked during processing (only Cancel active)
v2.6.1
2026-02
Fixed Faster-Whisper-XXL extraction error (solid 7z archives)
Generate button now disabled during download/extraction
Added download size and extraction duration hint
v2.6.0
2026-02
Faster-Whisper-XXL: AI speech recognition with VAD for accurate timestamps
Automatic backend selection: Faster-Whisper preferred, whisper.cpp as fallback
Language-specific whisper models: Chinese (Belle-large-v3) and Japanese (Kotoba v2.0)
Specialized model download: One-click install from HuggingFace
Sharpen/Denoise filters now work in Copy mode (auto-transcode)
Improved file size estimation for hardware encoders (NVENC, QSV, AMF)
Progress bar fix: Overall progress now updates correctly for single files
Stability: Process cleanup on cancel, thread-safe progress tracking
v2.4.0
2026-02
A/B Comparison view with frame-by-frame quality check
VMAF quality analysis (integrated libvmaf scoring)
Test encode: 30-second preview before full encoding
ASS/SSA subtitle style editor (fonts, colors, alignment)
Scene detection with automatic chapter generation
HDR tonemapping and deinterlace filter support
Custom filter chain builder (20+ FFmpeg filters)
Legacy format detection with modernization presets
Dolby Vision profile detection and conversion
Improved progress bar with real-time percentage
v2.3.2
2026-02
Fixed GPU detection on multi-GPU systems (dGPU now preferred over iGPU)
All GPUs detected: NVENC, QSV, and AMF tested independently
Unavailable encoders are now grayed out in the dropdown
GPU info shows all available encoders at a glance
v2.3.0
2026-02
Intro detection with audio fingerprinting
Credits detection (silence analysis)
Batch OCR for all files in queue
In-app update mechanism
Automatic signature updates for intro detection
v2.2.20
2026-02
Batch OCR for multiple files at once
Interactive spell check in the OCR dialog
Watch Folder monitoring
Improved VobSub timing correction
v2.2.0
2026-01
GPU encoding: NVENC, AMF & QSV support
New crop editor with preview
Encoding presets (custom & predefined)
Encoding history with statistics
v2.0.0
2025-12
Completely new UI with dark theme
Integrated subtitle editor with waveform
Native MKV parsing
Drag & drop import
2026-06
Redesigned toolbar: tidy groups instead of a cramped icon strip, rare tools tucked into a "" menu, and watch-folder actions now sit next to the watch toggle — readable in both dark and light mode
Reworked track selection rules: clear groups (Audio / Subtitles / Burn-in), plain language instead of cryptic terms, a single dropdown for forced subtitles — and the "forced subs matching the audio language" mapping now actually works (e.g. Japanese audio German forced subtitles for signs/songs)
Per-preset file naming: each profile remembers its own naming scheme, so a "Series" profile names episodes differently from a "Movies" profile — including for files auto-encoded from watch folders
Settings that weren't being remembered (notifications, sound, taskbar flash, conflict action and more) now persist; the "parallel jobs" slider works again
The Command tab now shows the real, runnable ffmpeg command (including 2-pass) and refreshes automatically when opened
Surfaced hidden features: VAD/pacing fine-tuning and the diagnostics export now live in the Settings tab
Smaller fixes: only one "Support VidControl" prompt after batch encodes, a de-cluttered Queue tab, filled-in translation gaps, and presets no longer lose their track rules when saved
v2.9.30
2026-06
Watch folders: a new switch, Auto-encode detected files. With it on, anything dropped into a watched folder is encoded automatically — previously you had to start each file by hand. Off keeps the old behaviour: files just go into the queue and you start them yourself
Removed the Encoding report window that popped up after each batch. It blocked the start button until you clicked it away. The Windows notification stays, and the full summary is still in the Log tab
v2.9.29
2026-06
New: AI dubbing. VidControl can now voice your translated subtitles itself. It pulls the voices out of the video, clones one per speaker and builds a complete new audio track. The music and sound effects stay intact, and the finished track is added to the video as an extra audio track on your next encode
New: Dub Studio for hand-tuning. Every voice sits on its own lane on a timeline. You can listen to individual lines, move, shorten, raise or lower them and fade them in and out. If lines run into each other, one click on 'Resolve overlaps' pulls them apart. Single lines can also be regenerated
New: audition a voice first. Before the long run starts, pick a speaker, type any sentence and hear right away whether the cloned voice works
AI dubbing is optional and runs entirely locally on your graphics card. VidControl downloads what it needs by itself the first time
Clearer codec list: codecs that can't run in the current mode (e.g. VVC or AV1 under NVENC) now show greyed out with a reason instead of just being missing, so you can see why something isn't selectable
Fixed the dub-engine setup: the window now closes by itself after the download and continues straight on; the close button no longer hangs
v2.9.28
2026-06
Pulled before release due to a dub-engine setup soft-lock — all of its features ship in 2.9.29
v2.9.23
2026-05
NEW — Hybrid transcription: choose the timing model (segmentation & timestamps) and the text model (wording) separately. Perfect when a general model like large-v3 times speech well but a language-specialised model like Belle understands the language better — it gives you the good timing AND the good words. Directly fixes the case where Belle assigns single words multi-second cues
The two models run one after another, so even a single GPU under 16 GB only ever loads one at a time. A recommendation is shown based on your graphics card (VRAM) — but every choice is yours; nothing is hardcoded
The text model's words are aligned onto the timing model's segments by content (character alignment), not by the text model's untrusted timestamps
v2.9.22
2026-05
Fixed the readability re-segmentation so it actually fixes heavily fragmented subtitles (one word per cue — common with the Chinese model + word-by-word translation). A reading-speed gate had been blocking the very merges it needed; word-cues are now correctly merged back into full sentences
Added a timing step that extends merged cues into the following gap until they are comfortably readable (reading-speed target + minimum duration), never overlapping the next cue
Splitting now triggers only when a cue is on screen too long or has too much text to fit — not on reading speed, which more cuts can never improve
v2.9.21
2026-05
NEW — Context-aware translation: the LLM now sees the already-translated neighbouring cues (a rolling window) plus a few upcoming source lines, instead of translating each line in isolation. This is the biggest readability win for translated subtitles — terminology, names and register stay consistent across chunk boundaries, and sentences that span several cues read naturally
On by default (4 context lines); tunable via the TranslationContextSize setting (0 turns it off). Context lines are clearly marked "do not translate", so the index mapping and output are unchanged
v2.9.20
2026-05
NEW — Readability re-segmentation: a decoupled engine that re-cuts subtitles so the text reads naturally — merges fragmented cues, splits over-long/too-fast ones at the most natural boundary (sentence comma conjunction word), and rebalances lines, with separate CJK and Latin rules and proportional time redistribution
Runs automatically after translation (on by default, with a toggle in the translation dialog) — fixes the common case where translated German breaks at the original Chinese cue boundaries instead of natural German ones
Also available as a "Re-segment" button in the subtitle editor (Fix tab), reversible with Ctrl+Z; works on any subtitles — imported SRT, OCR output, external translations, batch repair
Translation dialog is now scrollable with the buttons pinned at the bottom, so the "Translate" button is always visible even on small or scaled displays
v2.9.19
2026-05
NEW — Target-VMAF auto-CRF: VidControl sample-encodes a few representative segments and binary-searches the highest CRF (smallest file) that still reaches your quality target (e.g. VMAF 95). One button in Advanced Video settings opens a live probe dialog and applies the chosen CRF to your preset
VMAF analysis fixed: an unescaped Windows drive-letter colon in the libvmaf filtergraph meant VMAF never actually ran; also made resolution-robust via scale2ref
Fixed a startup crash in 2.9.18: duplicate localization keys (left over from the i18n waves) threw before logging initialized, so the app exited with no error dialog and no log
Diagnostics export now redacts webhook URLs (not just keys/tokens)
Subtitle editor: pacing now operates on the correct file if you switch tracks during the audio analysis; removed double event-subscriptions on track load; QC auto-fixes pick the correct neighbour cue when two cues share a start time
Settings are saved thread-safely — structural list mutations during parallel encodes can no longer make a debounced save silently fail
v2.9.18
2026-05
Command palette (Ctrl+Shift+P): fuzzy-search every command, preset and recent file — any of the ~119 actions reachable in two keystrokes
Universal undo/redo in the subtitle editor (Ctrl+Z / Ctrl+Y) across all ~28 mutating operations — batch and AI edits are now safely reversible
Subtitle QC linter with one-click "fix all" (Netflix/EBU profiles, dedicated CJK reading-speed & line-length limits): reading speed, line length/count, min/max duration, overlaps and gaps
Snap-to-silence activated: cue edges now align to real detected audio gaps in both the editor and the Whisper flow; waveform peaks are cached to disk so reopening is instant
Crash-safe settings: atomic writes (no more truncated settings.json), debounced auto-save, and a redacted diagnostics .zip export for bug reports
Encoding: HDR10 static metadata (MaxCLL + mastering-display) is now preserved on libx265 output, AV1 film-grain synthesis, an encoder -tune selector, sharper Lanczos downscaling, +faststart for MP4, NVENC spatial/temporal AQ, and a VMAF analysis fix that made it work on Windows at all
Translation: pinned term overrides ("this name always becomes that") are now enforced after the LLM, a free-text style/register instructions field, and a higher Whisper decode-quality floor
v2.9.17
2026-05
FFmpeg setup dialog on first run: auto-download from BtbN (~80 MB), point at existing ffmpeg.exe, or skip for now — no more "where do I get FFmpeg" wall for new users
Extended FFmpeg detection chain: settings path tools/ffmpeg/ tools/ffmpeg/bin/ PATH common install locations
FFmpegProcessManager hot-swappable path — after auto-download, encoding becomes available without an app restart
CLAUDE.md updated: FFmpeg is deliberately not bundled to keep VidControl decoupled from FFmpeg's monthly release cycle and avoid LGPL/GPL attribution overhead
v2.9.16
2026-05
Video codec dropdown now filters by hardware capabilities — codecs your NVENC/AMF/QSV can't actually encode (e.g. VP9 on any GPU, AV1 without Ada/RDNA3/Arc) are no longer selectable, preventing silent CPU fallbacks
ASS/SSA subtitle burn: embedded font/size/outline from the source file is now honored by default instead of being overridden with Arial (new override checkbox if you want the old behavior)
HEVC 10-bit on Pascal NVENC: detects the GPU can't encode 10-bit HEVC and converts to NV12 automatically instead of silently dropping to 8-bit
Settings loader now retries on file locks (OneDrive sync, antivirus, second VidControl instance) instead of falling back to defaults and silently losing user settings
Subtitle editor distinguishes save errors: read-only file vs. locked by another process vs. generic — clearer feedback
Whisper process kill uses a retry loop on cancel — no more zombie processes
UTF-16 subtitle files without BOM are now detected (previously garbled text)
Custom FFmpeg audio args now sanitized the same way video args are
FilePath null validation in the encoding pipeline — explicit error instead of cryptic "no such file" from FFmpeg
Resegmenter: crash guard for files with 1 entry
v2.9.15
2026-05
Complete English localization — every dialog, status message, queue state, and error text fully translated (around 400 strings localized across two waves)
OCR default language now follows the UI language (English UI English OCR; previously always German regardless of UI)
First-run language derived from your OS culture instead of forcing English
Output folder selector promoted to the main toolbar with browse + one-click „use source folder" reset (was previously buried in Settings)
Whisper VAD parameters (threshold, min silence, max speech per segment) exposed in Settings for power users
Belle Chinese model made usable — Silero VAD pre-segmentation forces sub-cues 12s instead of 30-60s blocks
CJK-aware post-splitting: Chinese subtitles split at with character-proportional time distribution
Whisper hallucination detection: phonetic-floor heuristic auto-drops garbage cues under 0.40s
Pacing review dialog with two tabs (hallucinations + merges) and audio preview per cue
Update system fixed — was reporting 2.7.4 as latest despite newer 2.9.x releases
Ctrl+H opens Find & Replace in the subtitle editor; tooltips added for video player and sync point actions
v2.7.4
2026-03
EAC3/Atmos audio in M4V output: FFmpeg muxer fix prevents exit code -22
Intro detection retroactive: Already loaded files are scanned when enabling the feature
Fixed pipe deadlocks in 5 services (DoviTool, AutoSync, PostProcessing etc.)
Crash safety: fixed async-void crash in GPU monitor and update dialog
Thread safety & resource leak fixes across 22 files
OCR spell correction: Replacements now context-aware per word
PropertyChanged notifications for file size, bitrate, speed, ETA
v2.7.3
2026-02
Remove scene tags: Option to automatically strip release tags from filenames (e.g. German AC3 480p WEBRip XviD-FND)
Queue context menu: Open output folder, open source file, move output file
PDB files are no longer generated in release builds
v2.7.2
2026-02
Custom titlebar: Toolbar is now the title bar — saves one row of height
Min/Max/Close buttons clearly visible in purple toolbar (dark theme fix)
Status bar at bottom doubles as drag area for window movement
v2.7.1
2026-02
Minor bugfixes and stability improvements
v2.7.0
2026-02
Settings decluttered: Progressive disclosure with collapsible expanders
Advanced Video, Custom Filters, Expert Options, Post-Processing now collapsed by default
Track Selection Rules: Basic options visible, advanced options in sub-expander
~40% less visual noise in Settings — no features removed, just hidden for power users
Fixed: Auto-update check did not detect new versions due to HTTP caching — please update manually this one time
v2.6.5
2026-02
Whisper Review: Spell check & consistency review after transcription
Interactive spell check with Hunspell (same as OCR review)
Consistency checks: duplicate entries, high CPS, short duration, inconsistent spellings
Automatic language mapping (Whisper Hunspell dictionaries)
v2.6.2
2026-02
Whisper dialog: All inputs locked during processing (only Cancel active)
v2.6.1
2026-02
Fixed Faster-Whisper-XXL extraction error (solid 7z archives)
Generate button now disabled during download/extraction
Added download size and extraction duration hint
v2.6.0
2026-02
Faster-Whisper-XXL: AI speech recognition with VAD for accurate timestamps
Automatic backend selection: Faster-Whisper preferred, whisper.cpp as fallback
Language-specific whisper models: Chinese (Belle-large-v3) and Japanese (Kotoba v2.0)
Specialized model download: One-click install from HuggingFace
Sharpen/Denoise filters now work in Copy mode (auto-transcode)
Improved file size estimation for hardware encoders (NVENC, QSV, AMF)
Progress bar fix: Overall progress now updates correctly for single files
Stability: Process cleanup on cancel, thread-safe progress tracking
v2.4.0
2026-02
A/B Comparison view with frame-by-frame quality check
VMAF quality analysis (integrated libvmaf scoring)
Test encode: 30-second preview before full encoding
ASS/SSA subtitle style editor (fonts, colors, alignment)
Scene detection with automatic chapter generation
HDR tonemapping and deinterlace filter support
Custom filter chain builder (20+ FFmpeg filters)
Legacy format detection with modernization presets
Dolby Vision profile detection and conversion
Improved progress bar with real-time percentage
v2.3.2
2026-02
Fixed GPU detection on multi-GPU systems (dGPU now preferred over iGPU)
All GPUs detected: NVENC, QSV, and AMF tested independently
Unavailable encoders are now grayed out in the dropdown
GPU info shows all available encoders at a glance
v2.3.0
2026-02
Intro detection with audio fingerprinting
Credits detection (silence analysis)
Batch OCR for all files in queue
In-app update mechanism
Automatic signature updates for intro detection
v2.2.20
2026-02
Batch OCR for multiple files at once
Interactive spell check in the OCR dialog
Watch Folder monitoring
Improved VobSub timing correction
v2.2.0
2026-01
GPU encoding: NVENC, AMF & QSV support
New crop editor with preview
Encoding presets (custom & predefined)
Encoding history with statistics
v2.0.0
2025-12
Completely new UI with dark theme
Integrated subtitle editor with waveform
Native MKV parsing
Drag & drop import