updates.json

{
    "1.0.0": "Initial release",
    "1.0.1": "Added updates menu",
    "1.0.2": "Fixed some issues causing infinite loading screens",
    "1.0.3": "- Improved the UI - reorganized the buttons to group the letter-specific tasks to the top bar, and sequence-wide setting on the bottom bar\n- Added a 'Reset letter' button\n- Made it more clear which letter is being edited by highlighting it in red\n- Added an option in the settings for auto-playing the audio when it's generated\n- Fixed an issue where question marks in the input broke file saving",
    "1.0.4": "Bug fixes",
    "1.0.5": "Bug fixes",
    "1.0.6": "- Added a file logger for the front-end to aid with user support/debugging - app.log\n- Fixed audio not saving if filename-unsafe characters are in the input sequence\n- Reduced freedom of range on the pitch slider to reduce bad quality output\n- Fixed issue with auto-regen spamming the model with every slider tick and filling the screen with modals (all crashing if also making the app run out of vram)\n- Fixed issue with deleting newly generated files\n- Other small tweaks and fixes",
    "1.0.7": "- Made the amplify/flatten buttons additive instead of multiplicative (to avoid it getting out of hand)\n- Fixed the letter length/pacing sliders! (I could never reproduce the issue, but I'd be surprised if this didn't fix it)\n- Fixed a few small path issues, potentially also fixing a bug where temporary files were not getting cleaned\n- Small UI changes, including file record hover effect, and including the file name in the deletion prompt\n- Fixed newline characters in the prompt breaking file saving\n- Added the ability to specify (on a per-game basis) what directory the output files should be placed in, when saved",
    "1.0.8": "- Changed the default window size (you can always resize this by dragging the corner) and enlarged settings menu\n- Fixed bad default custom filepaths in settings that may have lead to some audio output issues (you will still need to adjust this if you've already loaded v1.0.7, as those values are now bound to your personal settings - remove the double 'resources/app' if you have that - just check that the path exists)",
    "1.0.9": "- Added audio post-processing (via ffmpeg, you need to install it to enable this), to allow audio settings such as: file format, silence padding, sampling rate, and bit depth\n- Bug fixes for the pitch/length editor, for when the input is changed (then 'Something went wrong' happens)\n- Reduced the rate of increase/decrease of the Amplify/Flatten buttons",
    "1.1.0": "- Ability to edit previously generated samples* Numerical inputs for sliders\n- Upgraded WaveGlow to a better version. Bigger and a bit slower, but better sounding output across the board\n- Re-implemented a big part of the letter lengths mechanism, fixing many issues in the process\n- Ability to edit multiple letters at once (ctrl+click multiple), moving pitch slider together as a group\n- Compiled using CUDA 11.0 (fix for 3000 series GPUs)\n- Shift-click 'Keep sample' to first bring up a file name prompt\n- Stopped disabling the 'Keep sample' button (useful for when outputting several audio post-processing versions)\n- Made clicking on the letter bring focus to it (instead of clicking the slider and accidentally changing its value)\n- Made the pitch editor taller\n- Show the values for the editor sliders in a local tooltip\n- Fixed bug with renaming files with explicit file format in the name\n- Fix mp3 audio post-processing output sometimes failing\n- Fixes to file re-naming if full stops were present in the input sequence\n- Several other internal changes, including tweaks for any potential memory leaks\n- Keyboard shortcuts, with a new 'Information' menu, containing this cheat sheet:\n left/right arrows: Move between letter focused\n SHIFT-left/right: multi-letter create selection range\n up/down move pitch up/down for the letter(s) selected\n CTRL+left/right arrows change the sequence-wide pacing\n CTRL+up/down increase/decrease buttons\n CTRL+SHIFT-up/down amplify/flatten buttons\n space bring focus to the input textarea\n Enter: Generate sample\n CTRL-S: Keep sample\n CTRL-SHIFT-S: Keep sample (but with rename prompt)\n Escape: close modals\n Y/N for prompt modals",
    "1.1.1": "- Various bug fixes on v1.1 features from feedback on the beta version\n- Added letter length support to the multi-letter editing mode\n- Enlarged the changelog/updates menu to fit the complete changelogs\n- Made amp/flat incr/decr buttons work only on selected range when multiple letters are selected\n- Fixed an issue with Amplify/flatten, increase/decrease buttons starting to overreact after the app has been in use for some time\n- Fixed issues with sluggishness of the app after some time of use has passed",
    "1.2.0": "- Added support for bespoke voice vocoder models - a fourth file in voice packs, a HiFi GAN model trained specifically for that voice with higher quality than either WaveGlow vocoders, at the speed of the 'quick-and-dirty' model. Denoted by a '🗲'' symbol next to the dropdown where available (To start with just F4: Nate)\n- Added support for models saved as FP16, meaning models from now on will only be half as big (in terms of file size)\n- Added an option to select which Vocoder you prefer to use (following feedback on v1.1 vocoder model change) - useful for deciding which to use for the best quality\n- Added an amplitude multiplier ffmpeg audio post-processing option\n- Added a search bar for the models selection panel\n- Fixed issue where keyboard shortcuts were active when typing in input fields\n- Added a specific error message to when the output directory for saving audio files does not exist\n- UI tweaks/changes, enlarged minimum size of modals, for extremely small window sizes (the window can be resized)\n- Fixed some occasional issues with file re-naming (and then editing)\n- Made the 'Enter' key act as a submit action when in the rename modal (and other prompt modals)",
    "1.2.1": "- Fixed reset button not resetting pacing, and the 'Reset Letter' doing it (swapped)\n- Fixed voice-specific sub-folders not being automatically created in the output directory",
    "1.2.2": "- Fixed bespoke HiFi models not showing up in the dropdown\n- Changed output directory settings inputs to a single input field, for the currently loaded game\n- Made clicking outside the editor de-select the letter(s)\n- Fixed case where dragging a slider for a letter outside a selected group also moved it without first de-selecting the group\n- Added a dark mode setting for the input text sequence, for accessibility",
    "1.2.3": "- Fixed some issues with ffmpeg use when overwriting files\n- Added support for a few new games (and game series): Fallout 76, Mass Effect, Half Life, The Witcher\n- Re-implemented the game selection menu, to support a long list of other supported games\n- Added app resilience against some extreme edge cases with badly formed models folder structure",
    "1.3.0": "- Added batch synthesis mode! (From a .csv or .txt file)\n- Added partial re-generation! For changing only parts of the text without losing editor progress\n- Fixed silence padding ffmpeg setting no longer working\n- Added an option to keep the same pace value on new text line generations (on by default)\n- Fixed the reset letter button not taking the sequence pacing into account\n- Added option to automatically switch to HiFi-GAN vocoders when loading a voice, if one is available (on by default)\n- Made the UI snappier (reduced the fade transition times)\n- Fixed some settings options always defaulting to true, on app start-up\n- Added even further resilience to ffmpeg not overwriting files\n- Changed some things under the hood to enable non-multispeaker FastPitch support (easier for others to add their own models to the app)\n- Added game support for: Cyberpunk 2077, Civilization\n- UI tweaks\n- Updated the Fallout 4 skin image to a higher res version",
    "1.3.1": "- Added settings to use per-game custom (user specified) models directories\n- Made the model loading automatically refresh with file changes (re-starting the app no longer needed when installing new models)\n- Various quality-of-life UI tweaks",
    "1.3.2": "- Added batched batch mode - so parallelizing voice generation (limited only by your VRAM)\n- Option to skip batch generating lines which have already been generated\n- Globally improved voice generation quality, where word spacing was too short (zero) by default\n- Various UI tweaks\n- Speed optimizations for the batch synthesis mode",
    "1.3.3": "- Fixed occasional issues with loading models from certain custom paths",
    "1.3.4": "- Improved error handling for when models cannot be found by the app",
    "1.3.5": "- Added 'Fast Mode' to batch synthesis (more parallelization)\n- Fixed a few small UI bugs\n- Fixed batch mode erroneous handling of no specified output path for .csv records",
    "1.3.6": "- Fixed special ellipsis character error\n- Fixed occasional model loading vocoder error\n- Fixed video links not opening in new windows\n- Added batch synthesis video link to the batch menu\n- Added EULA",
    "1.3.7": "- Added reset buttons for either settings, or paths, in the settings menu\n- Added a toggle option for automatic voice reloading on file changes (some people experienced issues with the app 'unloading' voice models with it turned on)\n- Made the 'Drag+drop' note change to indicate when the app is busy processing given files (visible when big files are used)\n- Reworked/fixed UI modal switching behaviour\n- Default the editor values to 0 if null to avoid edge case errors\n- Fixed the batch output directory not being created recursively on batch start\n- Disallowed null input for batch output folder\n- Added game support for: Deus Ex, Star Trek Armada II",
    "1.4.0": "- Added a speech-to-speech mode\n- Added a third party Plugin system. Developer reference: https://github.com/DanRuta/xVA-Synth/wiki/Plugins\n- Added numerical input field for pacing\n- Added internationalization support for the app (other languages for the app UI now supported, via the plugins system)\n- Added setting for app audio output device\n- Added secondary audio output device for Ctrl+clicking 'play' on audio\n- Added a settings option to disable .json output\n- Added an option to name files numerically sequentially (filename_0001.wav, etc)\n- Added a button on records to edit the audio files with a configurable external program (eg Audacity)\n- Added a setting option to allow keeping the same editor state while switching voice (transfer speaking style from voice A to voice B)\n- Added alt key modifier to the editor for making whole word selection range\n- Added hover tooltips to the records' utility buttons\n- Added the community guide link to the info menu\n- Stopped clearing editor selection on letter mis-click\n- Improved app UI on first-time use before anything is configured\n- Lowered the increments for pitch sliders, and raise/lower buttons\n- Fixed auto-regen behaviour/not being run for some editor buttons\n- Fixed amplify/flatten buttons behaviour\n- Fixed UI bug where the chrome was flickering when changing voice models\n- Made batch synth file format handling more stable\n- Made .txt batch synthesis lines use HiFi vocoder if one is available\n- Fixed 'Skip existing output' not working for .txt files added to batch synth\n- Remove final full-stops from the end of the file name when saving\n- Made the editor letters more readable via better CSS\n- Added setting for the fade opacity on the background image\n- Made UI modal switching snappier\n- Replaced the 'Open containing folder' icon with a better one\n- Various other small UI tweaks and fixes\n- Added user setting for font size in the text input\n- Added game series support for: Dragon Age, Command and Conquer, Star Trek: Bridge Commander\n- Fixed amplitude multiplier input not updating enabled/disabled on ffmpeg check\n- Fixed issue where attempting to generate empty input broke future gens\n- Shortened file names to alleviate windows max file name issues\n- Fixed the window looking squished at small window sizes\n- Re-disable editor controls when unselecting letters",
    "1.4.1": "- Made model inference faster\n- Fixed issue where the app needed restarting when uninstalling certain plugins\n- Fixed app crash on start, if no plugins folder exists\n- Added plugins event for pre/post batch synth\n- CTRL+A to select all the letters in the editor\n- ALT+CTRL+Left/Right arrows = adjust pacing only for selection range\n- Added game support for: Grand Theft Auto 4, Grand Theft Auto 5, Grand Theft Auto San Andreas, Bioshock, Resident Evil, Borderlands",
    "1.4.2": "- Added optional configurable multi-processing for ffmpeg output in batch mode, making things faster\n- Added Play/Edit actions to batch synth mode\n- Added several extra user feedback windows for errors in batch mode\n- Added generate-voice pre event for JavaScript plugins\n- Added hover tooltips for batch lines' transcripts and out path\n- Fixed UI not handling deleting voice records with already deleted files\n- Fixed UI bug in the info tab\n- Fixed the games list not updating when changing the models paths",
    "1.4.3": "- Hotfix for a bug introduced in v1.4.2 where ffmpeg no longer saved audio clips correctly\n- Added shift click selection creation in the editor",
    "2.0.0": "<b>Main app:</b>\n- Merged CPU and CPU+GPU - single installation for both, going forward\n- Added support for FastPitch 1.1 models\n- Added energy as another control vector for the sequence editor - affecting per-letter loudness\n- Added Nexus integration, for downloading/endorsing voices from any nexus page\n- Added 3D voice embeddings visualizer, for voice similarity search\n- Added app support for phoneme based input/editing, to allow inputting exact pronunciation \n- Added an ARPAbet dictionaries management menu, for automated explicit pronunciation insertion during synth\n- Pitch/duration(and now energy) editor v2.0. Re-implemented with a canvas rather than intricate, limited, and hacky css+html\n- Audio player v2.0, now showing the audio waveform above the pitch/duration editor\n- WaveGlow vocoder models are now optional\n- Made WaveGlow vocoders' paths configurable\n- Added setting+ability to play back only changed audio, on regenerations\n- App start-up is now much faster; Huge app backend re-write/re-organization\n- Added a tip-of-the-day menu to the app start-up, with plenty of tips for app usage\n- Added configurable toggleable error sound to errors+set task icon to red\n- Bundled ffmpeg installation into the app. Set 'on' by default.\n- Added option to pre-apply ffmpeg to the preview sample\n- Improved front-end startup loading scalability, making startup much faster. There were too many voices to load at once!\n- Re-implemented game asset file handling to use .json metadata files rather than filename metadata\n<b>Batch:</b>\n- Major batch mode speed gains from several tweaks, changes, and optimizations\n- Added pagination to the batch screen records, to get rid of UI lag\n- Added timing and ETA info display to batch mode\n- Stability changes for batch mode, especially fixing Fast mode\n- Fixed batch mode multi-processing not working correctly\n- Added setting for the CSV delimiter used in batch mode preprocessing\n- Added GPU VRAM usage meter to the batch menu\n- Added scalability settings toggles for batch importing grouping tasks (for 100k+ line files)\n- Added yellow paused colour to task icon when batch mode is paused\n- Removed server.log logging from batch mode, apart from errors\n- Fixes and improvements for batch edit action button\n- Fixed not automatically creating relative directories specified in out_path\n- Fixed backslashes in out_path columns breaking saving when editing a batch generated sample\n<b>Plugins:</b>\n- Added Nexus endorsements integration to plugins with a nexus link\n- Added .ini settings file registering to give plugins an easy way to have custom settings in the settings menu\n- Fixed plugins not being detected in some cases\n- Added a pluginsContext parameter to server calls, for custom front-end data access from back-end plugins\n- Added mp-output-audio plugins event for python code\n- Added plugins event for mid-synth to allow custom python control over the generated pitch/durs\n- Added batch-stop front-end event for plugins\n- Added error message to plugins init of JavaScript scripts, with error stack\n- Hoisted front-end functions to window scope for easy use/replacement in front-end plugins\n<b>Other:</b>\n- Added support for voice installation via the Steam workshop\n- Added copy+paste right click context menu to the dialogue input textbox\n- Added voice info hover tooltip next to voice titles\n- Added Discord status updating support, toggleable via settings\n- Added a search bar and sort buttons for the main page output voice records \n- Added windows filepicker buttons for all configurable file paths and directories\n- Added Ctrl+Shift+I console to production build, to aid user support debugging\n- Added support for other/expanded alphabets, starting with 'english_french_hepburn'\n- Added game support for: Overwatch, Team Fortress 2, Nier: Automata, The Outer Worlds, Humankind, Hyperdimension, Persona, Portal, Baldur's Gate\n- Added search bar for settings in the settings menu\n- Added a 'Delete All' button to delete all the files generated for a voice\n- Fixed first-time launch issue where output directory paths were not being correctly initialized\n- Fixed first-time launch issue where models were not found and the app had to be restarted\n- Cleaned up spammy error logging for when my server can't be reached for changelog/credits updates\n- Moved 'Auto regenerate' to the settings menu\n- Add a regex whitelist filter for text input, to stop unique chars from breaking the app\n- Improved keyboard navigation for rapid Ctrl+S usage\n- Added an option to also display the ffmpeg amplitude multiplier setting in the editor bar\n- Optimized asset images, reducing over 50% of total file size (>12Mb)\n- Upgraded internal pytorch version from 1.7.0+cu10 to 1.9+cu111\n- Added a /getAvailableVoices API endpoint for headless operations (this still needs the front-end to be running)\n- Fixed some issues with automatically numbered output filenames\n- Misc bug fixes\n- Misc UI fixes and improvements\n",
    "2.0.1": "- Fixed editing .json loading issues, and audio player sometimes not showing them\n- Prevented ARPAbet refresh on checkbox ticks for enabling/disabling words\n- Added warning note to mass ARPAbet enabling\n- Fixed 3D visualizer menu not finishing regeneration\n- Added Ctrl+R front-end refreshing, to avoid whole app re-starts if needed for FRONT-END issues",
    "2.0.2": "- Added better catching of errors from nexus use without being logged in\n- Fixed audio player no longer playing back audio after some time\n- UI tweaks/fixes",
    "2.0.3": "- Made automated ARPAbet replacements fast even with tens of thousands of active dictionary words\n- Made nexus downloads+installations reload files regardless of settings\n- Adding new ARPAbet words that exist in other dictionaries now prompt you for confirmation\n- Updated showcase video link to the v2.0 video",
    "2.0.4": "- Lots of fixes for speech-to-speech\n- Lazy-load ARPAbet dicts, asynchronously to handle huge dicts load lag better\n- Added option to output .json editor data in batch mode\n- Made editor ranges adaptive to generated values\n- Catch 3D visualiser error if filtering out to less than the minimum number of required voices\n- Added Discord server link\n- Fixed batch mode not always working with backslashes\n- Fixed plugins synth-line event pre_energy hook not registering; Other tweaks/fixes",
    "2.0.5": "- Hotfix for ARPAbet word enabling\n- Hotfix for batch .json not creating new directories first",
    "2.0.6": "- Made pitch/durs/energy toggleable for Speech-to-speech, with configurable strengths\n- Added multi-letter duration dragging to editor elements\n- Added support for ARPAbet variants(2), with numbered brackets\n- Made numerical pace value update with slider movements\n- Made the maximum number of fast mode parallelizations configurable\n- Added configurable max filename length (for maximum windows filepath length trimming); Fixed occurence of no trimming\n- Added tooltips to top right menu icons\n- Added Enable/Disable all buttons to Nexus games list\n- Added column sorting to the nexus voices menu\n- Added option to clear the text input box after generating audio\n- Added game support for Star Wars: Jedi Fallen Order",
    "2.1.0": "- Added support for voice variants\n- Big stability fixes for speech-to-speech, and slight boost in quality\n- Added pitch and tempo ffmpeg post-processing controls\n- Fixed inverted pacing in v2 models' batch mode use\n- Made batch mode .csv reading adapt to new delimiters, after user confirm\n- Add exception for .csv and .txt files in 'Clear directory first' batch mode option\n- Fixed path reset undefined issues\n- UI tweaks, misc small bug fixes and UI changes\n- Added game support for: GTA Vice City, Age of Empires",
    "2.1.1": "- Hotfix for s2s",
    "2.2.0": "- Adjusted main app UI bounds for extra space\n- Added a 'Jitter' button to main app, to randomly wiggle all/selected letters/phonemes\n- Added pitch amplify multiplier (Amplify/Flatten) column to batch mode\n- Made the Amplify/Flatten, Raise/Lower buttons take bigger steps\n- Fix for the audio player sometimes having issues with the pitch/pacing - Pull request credit: ionite\n- Added setting+option to only highlight models above a specific version, in the voices list panel\n- Added more robust server killing, on app exit\n- Added arpabet-replace pre/post event for plugins\n- Added plugins ability to cache data into the context for across multiple executions\n- Added google drive link to asset files for Steam users, where assets files can't be published\n- Added persistance to window position on desktop\n- Added strict \"\" searching to ARPAbet menu\n- Added .lip and .fuz renaming, for if the plugin is used\n- Better UI scaling for the ARPAbet menu\n- Stopped trimming spaces from beginning/end of text sequence to allow space starts/ends\n- Fixed some plugin context data writing not being used by the app\n- Misc fixes and tweaks\n- Stopped spam from occasional ARPAbet dict re-loading\n- Fixed issue where changes in the ARPAbet dictionaries would break repeat synths\n- Made batch mode output paths default to using ./ prefix, if no drive letter is present\n- Fixed ellipses preventing ARPAbet replacements\n- 'Metal Gear Solid' game series support in the app - credit: HappyPenguin\n- 'YouTube', 'Bloodborne' series support",
    "2.2.1": "- Fixed an issue with batch mode execution of v2 models \n- Fixed a bug with pitch amplification multiplier in batch mode \n- Added a toggle for batch mode to allow forcing numerical filenames \n- Allowed underscores in ARPAbet word replacements \n- UI tweaks \n- Added asset files for games with voice created by the community via xVATrainer: Assassin's Creed, Diablo, Dungeon Keeper, The Last of Us, Tomb Raider, Warcraft ",
    "2.3.0": "- Added xVATrainer-like metadata.csv support for batch mode input \n- Added prompt text into the out samples list rows \n- Added confirm window for choosing which values to reset \n- Added records search based on actual prompts in jsons \n- Added Save to CSV in the batch menu, to export all lines to a single file \n- Fixed batch input filename case sensitivity \n- Added more front-end error logging for easier user debugging help \n- Fixed errors with cuda/vram usage polling \n- Made alphabetical voice sorting not case sensitive \n- Fixed auto-regen for letter/phoneme length \n- Fixed pacing not being correctly reset \n- Prevented app from clearing batch records if broken file is added on top \n- Removed spaces/dots from start of filenames (security) \n- Fixed pitch/energy still being editable if clicking in the same place, even when toggled not visible \n- Show model json parse errors in a UI error modal \n- Misc fixes \n- Added community asset files for new game support: Black and White, Dragon Ball, Elden Ring, Halo, Legacy of Kain, Life is Strange, Shenmue, Starfox 64, The Legend of Heroes, The Wolf Among Us, Vampire Masquerade, Zelda; Added TTS datasets; Updated Starfield with official wallpaper ",
    "3.0.0": "<b>Headline changes:</b>\n- Added support for v3 models\n- Added multi-lingual support, with dropdown setting for the base language to process v3 TTS input as\n- Purged old speech-to-speech system. Replaced with new v3 voice conversion system, with new python-based mic recording\n- Added emotion sliders for: Angry, Happy, Sad, Surprised\n- Added rich text prompt editor with code-editor-like autocompletes for languages and ARPAbet\n- Added backslash lang control in sub-prompt components, allowing multiple languages per prompt\n- Added voice crafting system, and support for json-only voice models\n- Added style sliders, and management system\n- Added backwards-compatible audio 48khz super-resolution post-processing via nuwave2 diffusion model\n- Added deepfilternet2 'Clean-up' post-processing model alongside super-resolution\n<b>Post-processing:</b>\n- Added ffmpeg settings for noise reduction, with configurable parameters\n- Added SR and clean-up post-processing support to batch mode\n- Added de-clipping and de-clicking post-processing ffmpeg filters\n- Added de-essing post-processing filter\n<b>UI Changes:</b>\n- Upgrade from Electron v2 to v19\n- Added right click menu on output editor to copy ARPAbet sequence to clipboard\n- Added error modals for problems with Nexus API use, printing out the full response\n- More reactive Voice Conversion UI\n- Auto-insert spaces at start/end of text prompts\n- Made the automatic space padding configurable via user setting, and not affect the prompt box or output filename\n- Ensured temporary files are cleaned out after every synthesis\n- Added Tab autocompletion for selected text to ARPAbet via the g2p backend\n- Added explicit error message for unsupported ARPAbet symbol\n- Improved batch mode window design\n- Changed default ffmpeg Hz to 44100\n- Filtered out * asterisks from prompts\n- Cleared old sliders editor TTS data when voice conversion is used\n- Disabled useless json output for voice conversion samples\n- Added more obvious styling for disabled inputs\n- Removed batch fast mode\n- Removed now useless 'Automatically generate voice' Voice Conversion setting\n- Removed ffmpeg pre-apply setting, use as enabled going forward\n- Made sliders and checkboxes theme coloured\n- Added pagination to the output records, with user setting for pagination size\n- Filtered out non-audio files from the output records (Keeping only wav, mp3, ogg, opus, wma, xwm)\n- Removed game-specific app title changes. Only xVASynth going forward\n- Changed model details (i) info window opening to click, instead of hover to enable interactivity\n- Added v2 and v3 UI switching for ARPAbet symbol list displays\n- Added CC-BY-4.0 category\n- Added splash page with YouTube links, before the EULA\n- Hidden the Endorsements/Downloads counters from the Nexus Manage Repos menu, as they do not update\n<b>Under-the-hood:</b>\n- Re-built backend environments with Python 3.9\n- Add 'pfft' entries into the cmudict\n- Updated 3d emb visualizer to use better v3 embedding space model\n- Added ability to load multiple model instances for programmatic use of the app via local http\n- Added fallback model installation debugging mode for when models are loaded with missing asset files\n- Added the clean-up model as a pre-processing step for voice conversion\n<b>Fixes:</b>\n- Stopped start-up error messages being very quickly dismissed by the start-up process\n- Fixed Jitter button not doing anything for zero values\n- Fixed microphone recording not working for mono-only microphones\n- Stopped using cached g2p words if prediction failed (blank prediction)\n- Added more explicit and hopefully more robust device switching for synthesis (incl for super-resolution)\n- Improved/fixed error display/logging for batch mode\n- Fixed batch mode bug where pitch_amp only affects first batch line\n- Fixed batch mode output_path not being shown properly\n- Fixed editor issue where disabled sliders could still be moved if clicked in the right place\n- Fixed dashes being ignored in ARPAbet dictionary replacements\n- Fixed not all settings being reset \n- [Bunglepaws] Fixed manual ARPAbet input being broken in the text processing pipeline\n- Fixed 3D voice embeddings visualizer not displaying gender information correctly\n- Lots and lots of other misc tweaks, fixes, and polishing",
    "3.0.1": "- Fixed batch mode (tensor dimension error)\n- Fixed occasional broken sliders, especially for older voices\n- Added filtering to the languages dropdown by the languages listed in the voice json, if any\n- Fixed not being able to over-write files\n- Added batch mode resilience to csv files with extra spaces in the header values\n- Fixed batch 'Open Output' folder button\n- Fixed energy sliders broken on older voices\n- Fixed colons in words preventing dict phoneme replacements and falling back to g2p",
    "3.0.2": "- Fixed batch mode tensor dimension error (for real this time) \n- Added _ as an allowed symbol in custom ARPAbet dictionaries \n- Made auto-play enabled by default \n- Added versioning info into the saved json files. (App version, model version, model type) \n- Made 'Clean-up' enabled by default \n- [Bunglepaws] Fixed custom dictionaries use in the text pre-processing pipeline \n- [Bunglepaws] Fixed linux use of Electron's 'show in folder' \n- Fixed pitch sliders being outside the range for some models \n- Fixed the editor being broken sometimes on some older voices \n- Fixed batch mode output folder creation errors \n- Fixed nexus downloads not auto-creating missing models folders, and failing installations \n- Fixed inverted Raise/Lower editor buttons on v3 models \n- Fixed batch mode json outputting for v3 models ",
    "3.0.3": "- Potential fix for non-standard words not being processed by g2p \n- [Pendrokar] Added ability to adjust individual sliders without needing to explicitly grab the grabber \n- Fixed the sliders box being able to scroll past where the sliders end \n- Added over-ride setting for the allowable pitch range \n- Added Ctrl+Enter shortcut for re-generating a line \n- Made clicking anywhere dismiss the 'Convert to Phonemes' prompt box \n- Fixed issue where the Generate Voice button was enabled without a voice already loaded, when messing with the text prompt \n- Clicking in the text prompt box no longer clears editor selection \n- Fixed style embeddings not being removed from EDIT dropdown when changing voices \n- Fixed ARPAbet new/edited dictionary entries not accepting the new v3 symbols \n- Allowed phoneme swaps to not clear the slider values on re-gen"
}