Skip to content

updates.md

Latest commit

 

History

History
245 lines (193 loc) · 11.4 KB

updates.md

File metadata and controls

245 lines (193 loc) · 11.4 KB

Contrary Convertor — Updates

v1.9.0 (May 28, 2026)

Photo(s) → PDF / PDF → Photos

  • Convert any image (JPG, PNG, WEBP, BMP, TIFF, GIF, AVIF, HEIC) directly to a PDF page (embedded into A4 via pdf-lib)
  • Export PDF pages back to individual PNG images — one image per page, rendered via Puppeteer
  • Multi-image → single PDF: combine multiple photos into one PDF document

Windows Context Menu Integration

  • Installer registers "Convert with Contrary Convertor" in the Windows right-click menu for all file types
  • Opening a file via context menu automatically loads it in the app ready for conversion
  • Uninstaller cleanly removes the registry entries

Extended Archive Support (7zip-bin)

  • Full read/write for: ZIP, RAR, 7Z, TAR, TGZ, TBZ2, TXZ, BZ2, XZ, GZ
  • Extract-to-folder for all formats
  • Cross-format conversion (e.g. RAR → ZIP, 7Z → TAR)

Programming Language Files (new "code" category)

  • Supported: C, C++, Python, Rust, Julia, Kotlin, Nim, Dart, Go, Java, JavaScript, TypeScript, JAR, CSS, PHP, Ruby, Swift, Scala, Haskell, Lua, R, MATLAB, Perl, Shell, Batch, PowerShell, TOML, INI, CONF, LOG
  • Convert to PDF (syntax-highlighted via Puppeteer), HTML (dark theme), or TXT

Text-to-Speech (TTS)

  • Convert any text/code/document to MP3 using Microsoft Edge TTS (en-US-ChristopherNeural voice)
  • Auto-installs edge-tts and pydub on first use via Python
  • 192k bitrate MP3 output with normalization
  • Works on: TXT, MD, HTML, RTF, CSV, YAML, JSON, TOML, XML, and all code file types

Image Upscaling (2× / 4×)

  • Upscale any image 2× or 4× using sharp's Lanczos3 high-quality resampling
  • Non-destructive: saves as new file alongside the original

PPTX Conversions

  • PPTX → PDF, HTML, TXT, MD
  • No LibreOffice required — text extracted via native XML parsing

Expanded Markdown Output

  • PDF → MD, DOCX → MD, PPTX → MD, HTML → MD, TXT → MD
  • DOCX uses mammoth + turndown for clean Markdown from Word documents

Noise Removal — 3-Pass Pipeline

  • Pass 1: 6-filter FFmpeg chain: highpass → lowpass → afftdn (nr=35 nf=-25) → anlmdn → speechnorm
  • Pass 2: DeepFilterNet AI (unchanged)
  • New Pass 3: post-AI cleanup — afftdn + anlmdn + dynaudnorm + loudnorm for broadcast-ready output

Internal / Build

  • New npm dependencies: 7zip-bin, node-7z, marked, pptxgenjs, turndown
  • build/installer.nsh rewritten with context menu hooks and clean uninstall

v1.8.0 (May 28, 2026)

🎵 YouTube MP3 Fix

  • Fixed "invalid merge output format" error when downloading YouTube videos as MP3
  • Root cause: --merge-output-format mp3 is invalid — ffmpeg can only merge into container formats
  • Fix: switched to --extract-audio --audio-format mp3 --audio-quality 0 for audio downloads (VBR best)
  • Added --ffmpeg-location flag so yt-dlp uses the app-bundled ffmpeg binary reliably

🎧 Spotify Download Support

  • Paste any Spotify track, album, or playlist URL and download as MP3 @ 320k
  • Uses spotdl — auto-installed silently via pip on first use (leverages the existing Python auto-installer)
  • Format picker auto-locks to MP3 when a Spotify URL is detected
  • UI badge shows Spotify • via spotdl • MP3 • 320k on paste
  • Works for tracks, albums, playlists, and artist pages

v1.7.0 (May 22, 2026)

🔊 AI & DSP Background Noise Removal

  • New BG Noise Removal (2-pass AI+FFmpeg) post-processing option for audio and video files.
  • Pass 1 (DSP): Uses a double afftdn, anlmdn, and low-shelf equalizer filter chain in FFmpeg.
  • Pass 2 (AI): Uses DeepFilterNet AI network for advanced speech enhancement and noise reduction.
  • Silent Auto-Installer: Lazily installs Python 3.11 embeddable package and DeepFilterNet on Windows on first run (cached for future uses).
  • Graceful Fallbacks: Non-Windows platforms or environments without DeepFilterNet support fall back to the optimized Pass 1 DSP filters.
  • Supports single and bulk conversions, saving output as <name>_denoised.<ext>.

v1.6.0 (April 30, 2026)

🔖 PDF Watermark

  • New Watermark PDF conversion option — appears when input is a PDF
  • Customizable watermark text via text input
  • Opacity slider (0.00–1.00, default 0.15) for subtle or bold stamps
  • Text is drawn diagonally across the center of every page, auto-scaled to page size
  • Output saved as filename_watermarked.pdf in the same directory
  • Uses pdf-lib for fast, dependency-free PDF manipulation

v1.5.2 (April 30, 2026)

🐛 Critical Fix — Conversion ENOENT crash

  • Fixed spawn ffmpeg.exe ENOENT error that broke all video, audio, and image conversions in packaged builds
  • Root cause: ffmpeg-static resolved to a path inside app.asar, but OS cannot spawn executables from asar archives
  • Added getFFmpegPath() helper that swaps app.asar → app.asar.unpacked for the ffmpeg binary path
  • All 4 ffmpeg consumers patched: probeMedia, fixForPlatform, convertVideo, convertAudio

v1.5.0 (April 28, 2026)

📋 Clipboard Paste Support

  • Ctrl+V to paste files directly into the app
  • Works with copied files from Explorer and screenshots/images from clipboard
  • Clipboard images are saved to temp dir and loaded automatically

📦 Bulk Operations

  • Drop multiple files at once for batch conversion
  • Right-click the dropzone to open multi-file selection dialog
  • All files are processed sequentially with per-file progress reporting
  • Common output formats are computed from the intersection of all files' supported formats
  • Each result (success/failure) is individually tracked in history

🎵 Video → Audio Extraction

  • Videos can now be converted to WAV, OGG, FLAC, AAC, OPUS (not just MP3)
  • Full codec support: pcm_s24le for WAV, libvorbis for OGG, flac for FLAC, aac for AAC, libopus for OPUS
  • Proper format containers for each audio type

🔍 OCR for Images

  • Already built-in via extract-text format for images
  • Uses Tesseract.js for OCR on JPG, PNG, WebP, BMP, TIFF, GIF, AVIF
  • Real-time OCR progress reporting

v1.4.0 (April 28, 2026)

🔧 PDF Text Extraction — Complete Rewrite

  • All text is now extracted from PDFs using page-by-page getTextContent() via pdfjs
  • Image placeholders: Where images exist in the PDF, [Insert Image Here] is inserted in the output
  • Image detection uses the PDF operator list (paintImageXObject/paintJpegXObject ops)
  • Removed OCR dependency — pure text extraction, no Tesseract needed
  • Added pdf-lib fallback parser that reads raw PDF content streams (Tj/TJ operators) if pdf-parse fails
  • Works on all PDFs including encrypted/complex layouts

🎬 Video Conversion — Fixed

  • Added explicit .format() calls for every output container (mp4, mkv, mov, webm, avi, gif, mp3)
  • Added FLV and WMV output format support
  • Added -y overwrite flag to prevent ffmpeg hanging on existing files
  • Added -pix_fmt yuv420p and even-dimension padding for H.264 compatibility
  • Added -movflags +faststart for MP4 web streaming
  • Better error messages on ffmpeg failures
  • Added fallback encoder path for unknown formats

🖼️ Image Conversion — Hardened

  • Added failOnError: false to sharp — handles truncated/broken images gracefully
  • Added auto EXIF rotation via .rotate() — images no longer appear rotated after conversion

🎵 Audio Conversion — Fixed

  • Added -y overwrite flag to prevent ffmpeg hanging

Contrary Convertor — Updates (April 13, 2026)

🎨 UI Overhaul

  • Complete redesign with red + black gradient aesthetic
  • CRT scanline overlay with animated scroll effect
  • Glitch animation on logo text
  • Matrix binary text background in dropzone
  • Monospace typography (Share Tech Mono) across all UI
  • Pulsing red glow on logo hexagon

📂 Searchable Format Dropdown

  • Replaced format selection chips with a single searchable dropdown
  • Formats grouped by category: Image, Video, Audio, Data/Config, 3D, Extraction, Fix
  • Type to filter, scroll to browse — keyboard navigation support

🔧 New Conversion Formats

Fonts

  • TTF ↔ OTF conversion via opentype.js
  • Font metadata extraction (family, glyphs, version, designer, license)

Dev / Config

  • JSON ↔ YAML ↔ TOML ↔ .env bidirectional conversion
  • Auto-flattening for nested objects → .env keys

3D Models

  • OBJ → GLB (via obj2gltf)
  • GLB → OBJ (custom mesh parser with vertex + face extraction)
  • FBX → GLB (via fbx2gltf binary, if available)

Game Archives

  • .pak — Unreal Engine 4 (v1–v4) and UE5 (v11+) extraction
  • .rpf — Rockstar RAGE (GTA V / RDR2) best-effort extraction
  • .wad — Doom IWAD/PWAD lump extraction
  • .obb — Android OBB (ZIP-based) extraction

Text / OCR Extraction

  • Extract text from PDF, DOCX, XLSX, HTML, and any text file
  • OCR from images — JPG, PNG, WebP, BMP, TIFF via Tesseract.js
  • Extract images from PDF — renders each page as PNG
  • Extract fonts from PDF — lists font names + extracts embedded TTF/OTF files

PDF Improvements

  • PDF → TXT now uses pdf-parse for actual text extraction (with pdf-lib fallback)

⚙️ Settings Panel

  • Gear icon (⚙) in top-right titlebar
  • Slide-in panel showing:
    • Developer: Ayush.ue5
    • GitHub: github.com/AroseEditor/ (hyperlink, opens in browser)
    • Download Threads slider (1–16 parallel connections)
    • yt-dlp Quality preset (Best/Lossless, 1080p, 720p, 480p, 360p, Audio Only)

⬇️ URL Download Mode

  • "Want to download a link?" clickable text below dropzone
  • Switches UI to URL input bar — paste any link and download
  • Source detection — shows badges identifying the source (YouTube, Instagram, Telegram, etc.), download method (yt-dlp / Direct), and quality preset
  • Supported sources via yt-dlp (auto-downloaded on first use):
    • YouTube, Instagram, TikTok, Twitter/X, Facebook, Telegram, Twitch, Reddit, Vimeo, SoundCloud
  • Direct link downloads — multi-threaded with configurable parallel connections
  • Progress display — real-time percent, MB/s speed, downloaded/total bytes
  • Cancel button to abort any download
  • Quality presets applied to yt-dlp: lossless, 1080p, 720p, 480p, 360p, audio-only
  • Output format: MP4 for video, MP3 for audio-only

🖼️ AI Background Removal

  • Drop any image → select "REMOVE-BG — AI background removal" from the format dropdown
  • Auto-detects subjects using on-device AI (ONNX/WASM via @imgly/background-removal-node)
  • Paint/Erase mask editor with canvas overlay:
    • 🖌️ Brush — paint over areas to keep
    • 🧹 Eraser — erase areas from mask
    • Adjustable brush size slider (5–80px)
    • Crosshair cursor when painting on canvas
  • Apply & Save button — loading animation while processing
  • Saves as originalname_removedbg.png (transparent PNG)
  • AI model downloads ~50MB on first use (cached after)

🔧 Auto-Install Dependencies (No Manual Setup)

  • No more "Install Node.js" error — app auto-downloads portable Node.js (~30MB) on first launch
  • Downloads from nodejs.org to %APPDATA%/contra-conv/portable-node/
  • Uses that portable npm to install all conversion libraries
  • Cached — only downloads once, reused on all future launches
  • isFirstLaunch.txt flag tracks installation state (True = installed, False/missing = install)

🏗️ Build & Packaging

  • GitHub Actions workflow for macOS builds (DMG)
  • Linux AppImage build support
  • All new dependencies added to package.json