Repository navigation
Releases
FFmpeg Forge 2.3.0
FFmpeg Forge 2.3.0
New feature: Rip DVDs or Blu-rays
A new "Rip DVD/BD" button in the toolbar allows you to copy a DVD or Blu-ray into the video or audio format of your choice, using the FFmpeg version provided in the bin\ directory.
Supported sources
A DVD or Blu-ray drive
A VIDEO_TS (DVD) or BDMV (Blu-ray) folder
An ISO image
Disc reading
Lists titles (for DVDs) or playlists (for Blu-rays), including duration, chapters, video, audio tracks, and subtitles (with language information).
The longest title is selected automatically. You can select multiple titles; they will be ripped sequentially.
Choose specific audio tracks and subtitles for each title.
Titles shorter than one minute are hidden by default (this option can be disabled). Output Formats
Video: MKV (preserves everything; recommended), MP4, MOV, WebM
Audio-only: FLAC, MP3, M4A (AAC), WAV, MKA
Video
Direct copy (no re-encoding)
H.264 (x264), H.265 10-bit (x265), AV1 10-bit (SVT-AV1)
NVIDIA NVENC: H.264, H.265 10-bit, AV1 (RTX 40 / 50 series)
Settings: quality, speed, maximum resolution (1080p, 720p, 576p, 480p), deinterlacing (automatic for DVDs)
Audio
Copy, AAC, Dolby Digital (AC-3), Dolby Digital Plus (E-AC-3), FLAC, Opus, MP3
Adjustable bitrate and downmixing to stereo
In copy mode, tracks incompatible with the chosen output format are automatically converted (e.g., LPCM is converted to lossless FLAC when using the MKV container).
Subtitles
Preserved (in MKV), removed, or burned into the video image
Other
Chapters are preserved. For Blu-rays, durations, chapters, and languages are read directly from the playlists.
Progress bar with estimated time remaining, "Cancel" button, and detailed log.
Ripped files can be added to the main list for subsequent conversion or enhancement (RazorNext, VisionNext, noise reduction, etc.).
Rip settings are saved for future use.
Limitations
Copy-protected discs (CSS for DVDs; AACS/BD+ for Blu-rays) are not supported. The software will detect and flag them. FFmpeg Forge is designed for ripping your own personal videos, home-authored discs, and unprotected media. Summary of changes in version 2.2.x
AI Upscale: Choose between the VisionNext and RazorNext (RelayVSR) models for 4x video super-resolution, delivering stable detail from frame to frame. Also works on still images.
Integrated engine: One-click installation; no "Visual Enhancer" required; supports 1x to 4x scaling factors.
Visual Enhancer engine: Supports 1x, 2x, 4x, and 8x scaling factors, with adjustable GPU memory settings.
Installers (RazorNext, VisionNext, SeedVR2, QTGMC, AI runtime): Resolved "Failed to hardlink files" warnings when using external drives (exFAT/FAT file systems).
System requirements
Windows 10 / 11 (64-bit), .NET Framework 4.x (included with Windows)
DVD / Blu-ray ripping: Requires an optical drive, a VIDEO_TS / BDMV folder, or an ISO image.
NVENC encoding: NVIDIA graphics card (AV1 support requires RTX 40 / 50 series).
Integrated RazorNext: NVIDIA graphics card with 16–24 GB of VRAM recommended.
FFmpeg Forge 2.1.6
FFmpeg Forge 2.1.6
FFmpeg Forge 2.1.6 now available on Linux and MacOS
FFmpeg Forge 2.1.6 brings a brand new cinematic design, SDR to HDR conversion (HDR10 and Dolby Vision 8.1), and new audio track tools requested by users: change only the audio of a video, or add audio tracks, without re encoding the picture.
New cinematic design
New header with original artwork (film reel, golden light streaks), a bigger gold logo, "FFmpeg FORGE" in large letters and a gold "VERSION 2.1.6" badge.
FFmpeg status pill in the header: a magnifier while detecting, a green check when FFmpeg is found, a red warning when it is missing. "FFmpeg location…" now has its own button with a folder icon.
Every toolbar button has an icon: Add files, Add folder, Remove, Clear all, Show command, Filter library.
New drop zone: film strip artwork, big logo, "Drop your files here", and coloured feature badges (Video, Audio, Image, AI Denoise, AI Upscale, Dolby Atmos, DSD).
New Activity Log: one line per message with a status icon (info, success, warning, error), the time and the message, details dimmed. New "Clear log" button. Double click or Ctrl+C copies a line, right click copies everything.
Gold icons on every settings section, icons on the VIDEO, AUDIO and IMAGE buttons, taller file list rows.
Status bar with a coloured dot (green ready, gold working, red error), Open folder and Cancel with icons, and a new blue to gold CONVERT button.
Dark title bar on Windows 10 and Windows 11.
SDR to HDR (Video and Image)
New "SDR → HDR" option with a choice of peak brightness (600, 1000, 1500, 2000 or 4000 nits).
HDR10: H.265 10 bit, BT.2020 PQ, HDR10 static metadata (mastering display and content light level). VP9 or AV1 for WebM.
Dolby Vision 8.1: the same HDR10 picture plus Dolby Vision dynamic metadata, measured on every frame of your video and grouped by shot. It plays as Dolby Vision on Dolby Vision screens and as HDR10 everywhere else. MP4, M4V, MOV or MKV. Made with dovi_tool (MIT licence), included next to FFmpegForge.exe.
HDR pictures: PNG 16 bit with the HDR tag (cICP) or AVIF 10 bit, BT.2020 PQ. Shown in HDR by Chrome, Edge, Windows Photos with HDR turned on, and recent phones.
How it works: the SDR picture is placed in HDR with SDR white at 203 nits (the BT.2408 reference white), then the highlights are expanded up to the chosen peak. Only the brightness is expanded, colours keep their exact hue.
An HDR screen is needed to see the result: on an SDR screen, HDR files look dim or flat. Files that are already HDR are not converted again.
Audio tracks without re encoding the video
New "AUDIO TRACKS (VIDEO UNTOUCHED)" section in the Video category, as requested by users.
"Keep the video as is": the picture is copied bit for bit, same quality, very fast. Only the audio changes, for example Opus to AAC or AC3 (choose the codec in Audio settings).
Original audio tracks: convert them, keep them as they are, or remove them. Every audio track is processed, not only the first one.
Add audio tracks from audio files or from other videos, with language (guessed from the file name: _fr, VF, english…), title, delay in milliseconds to resynchronise a track, and default track.
Added tracks are copied as they are when the output format accepts them, otherwise converted automatically (the log says so). Chapters and metadata are kept, and subtitles too in MKV.
Also included
QTGMC deinterlacing (from 2.1.5) and the NVIDIA RTX AI audio sync fix (from 2.1.0.1).
Cover / thumbnail now works on Dolby Vision and HDR files too.
Support FFmpegForge
Enjoying FFmpegForge? If you find it useful, please consider stars the project on GitHub! It helps more people discover this free software and supports its continued development.
FFmpeg Forge 2.1.5
✨ FFmpeg Forge 2.1.5 for Windows
✨ [NEW] FFmpeg Forge 2.1.5 for MacOS
✨ [NEW] FFmpeg Forge 2.1.5 for LINUX
FFmpeg Forge 2.1.5 brings QTGMC deinterlacing for interlaced videos (DV, DVD, TV recordings, 1080i camcorders) and, for the first time, a portable edition for macOS, on Apple Silicon and Intel Macs.
🎞️ Deinterlacing with QTGMC (Windows and macOS)
✦ New "Deinterlacing" option in Video settings, with three choices: Off, QTGMC double rate (50p / 60p) and QTGMC same rate (25p / 30p).
🎬 Double rate turns every field into a frame: 25i becomes 50p and 29.97i becomes 59.94p, for the smoothest motion. Same rate gives one frame per original frame: 25i becomes 25p and 29.97i becomes 29.97p.
🧭 Field order: Auto uses the order written in the file (top or bottom field first). If motion judders back and forth, force TFF or BFF. Forcing an order also deinterlaces files that are not flagged as interlaced.
✅ Only interlaced videos are processed. Progressive videos are left as they are, and the log says so.
🧠 QTGMC is QTempGaussMC from vsjetpack, running on VapourSynth R81 with its plugins (MVTools, znedi3, zsmooth, resize2, akarin, vszip, BestSource). It runs first, before every other effect, so the rest of the chain gets clean progressive frames. Trim, audio, PNG/JPG SEQUENCE, cover and every format keep working.
📥 "Install QTGMC (VapourSynth)…" installs everything once: uv, Python 3.12, VapourSynth and the plugins at tested versions, about 150 MB. The installer checks that QTGMC loads before it finishes.
🛟 Until QTGMC is installed, FFmpeg's bwdif deinterlacer is used with the same settings, and the log says so clearly.
🐢 QTGMC is slow: it is the best quality option, not the fastest. On a plain conversion, bwdif is much faster.
🍎 New: FFmpeg Forge for macOS (portable)
💻 A native Mac application with the same blue and gold design and the same layout as the Windows version: files in the middle, every setting in one column on the right, CONVERT at the bottom.
📦 Portable: everything is inside FFmpeg Forge.app, FFmpeg included. Copy it to Applications, a USB stick or an external drive. Requires macOS 12 Monterey or later.
✅ Included on macOS:
Video, audio and image conversion with the main formats and presets of the Windows version (MP4, MKV, MOV ProRes, DNxHR, WebM, GIF, MP3, FLAC, WAV, PNG, JPG, WEBP…)
Quality, encoding speed, resolution, frame rate, rotation, trim and destination
Filters: preset library and your own FFmpeg filters
Compression assistant: analysis, real measurement on 3 samples and Optimized compression
Cover / thumbnail, PNG/JPG SEQUENCE and QTGMC deinterlacing
Apple VideoToolbox hardware encoders (H.264 and H.265)
Drag and drop of files and folders, remaining time for each file
⏳ Not available on macOS yet: NVIDIA RTX Video, NVENC and every CUDA feature (they need an NVIDIA GPU), Denoiser IA, Python / ONNX AI, VisionNext, SeedVR2, Scale / OptiScaler, Spatial Audio and DSD export. The AI features on the Apple GPU are planned for a later version.
🧩 QTGMC on macOS needs macOS 15 Sequoia or later (the VapourSynth plugins for Mac are published for macOS 15). On older versions, bwdif is used.
🔧 Improvements (Windows)
🎯 NTSC frame rates (29.97, 59.94, 23.976) are now written as exact fractions (30000/1001, 60000/1001…) when a frame rate is set.
🛡️ Unchanged and preserved (Windows)
The compression assistant, cover / thumbnail, NVIDIA RTX Video, PNG/JPG SEQUENCE, the live before / after comparison, every preset, codec and format, the AI features and your saved settings work exactly as before.
📦 Downloads
🪟 Windows
🎒 FFmpegForge-2.1.5-Portable.rar (about 206 MB): everything included, FFmpeg Forge 2.1.5 with FFmpeg 9.0.2 full and ForgeRTX. Extract the whole archive anywhere and run FFmpegForge.exe.
🪶 FFmpegForge-2.1.5.rar (about 0.7 MB): normal edition for your own FFmpeg (C:\FFmpeg\bin or next to the .exe).
🧩 Denoiser IA models (HAT L, DAT, SPAN): use FFmpegForge-2.0.0-Denoiser-Models.rar, unchanged. Extract it next to FFmpegForge.exe (models folder).
🍎 macOS
🚀 FFmpegForge-2.1.5-macOS-AppleSilicon.zip (about 140 MB): for Macs with M1, M2, M3 or M4, with FFmpeg 7.1.
🖥️ FFmpegForge-2.1.5-macOS-Intel.zip (about 164 MB): for Intel Macs, with FFmpeg 6.1.1.
🔓 First start on Mac: the app is not signed by Apple. Right click FFmpeg Forge.app, choose Open, then Open again. On macOS 15 and later, if it is blocked, go to System Settings, Privacy & Security, and click "Open Anyway". If macOS says the app is damaged, run this in Terminal: xattr -dr com.apple.quarantine "/path/to/FFmpeg Forge.app"
ℹ️ VapourSynth, vsjetpack and the plugins are not included in the archives: the installer downloads them from PyPI.
Enjoying FFmpegForge? If you find it useful, please consider stars the project on GitHub! It helps more people discover this free software and supports its continued development.
FFmpeg Forge 2.1.0.1 (Small Update)
A small update to FFmpeg Forge 2.1.0 that fixes the audio / picture desynchronization some users noticed after an export with the NVIDIA RTX AI option (RTX Video Super Resolution and RTX Video HDR, built in NVIDIA driver engine).
🎯 What was wrong
🎞️ With NVIDIA RTX AI, the picture goes through the NVIDIA driver as raw frames, without timestamps, while the sound is taken directly from the source file. When the video and the audio of the source do not start at exactly the same moment, that offset was lost: the sound ended up early or late for the whole video.
📼 It mainly happened with TS, M2TS and MTS files (camcorders, TV recordings, Blu-ray), with MKV files, with MP4 / MOV files that contain an edit list (phones, cameras, editing software), and when Trim was used. In our tests, a TS recording came out with the sound 0.38 s early.
✅ What is fixed
⏱️ Both sides of the RTX export now use the same time reference taken from the source file: the first frame of the RTX video and the first instant of the sound are the same moment of the original, whatever the start offsets of the streams.
✂️ Trim is now applied on the real timestamps of the file, the same way for the picture and for the sound, so trimmed RTX exports stay in sync too.
🔇 If the sound starts after the picture in the source, the right amount of silence is added at the beginning instead of pulling the sound forward. The sound also stays locked to its timestamps over the whole video, so it cannot drift on long or variable frame rate videos.
🎧 Audio "Stream copy" (no re encoding) keeps the right offset as well.
Enjoying FFmpegForge? If you find it useful, please consider stars the project on GitHub! It helps more people discover this free software and supports its continued development.
FFmpeg Forge 2.1
FFmpeg Forge 2.1.0
Thanks to the feedback received on the VideoHelp forum, FFmpeg Forge 2.1 helps beginners make their videos smaller without having to understand compression settings, while keeping every option for advanced users.
🧭 Compression assistant
✦ After you import a video, FFmpeg Forge checks whether it could be made much smaller. A large file alone is never enough: the assistant compares the real density of the video stream (bits per pixel and per frame, from ffprobe) with what a modern encoder needs for the same resolution and frame rate.
💡 When a worthwhile reduction is expected, a small card appears above the file list, with no pop up: current size, estimated size, estimated reduction, recommended video codec, container, audio and conversion time. Three buttons: Apply recommended settings, Measure size and time, Ignore.
No recommendation is shown for videos that are already efficiently compressed (HEVC, AV1, VP9 at a sensible bitrate), for HDR videos, for very short clips or for files that cannot be analysed.
🎛️ Recommended settings: H.265 (x265) at "Very high" quality, or H.265 NVIDIA NVENC for long videos when a working NVIDIA GPU is detected, AV1 for WebM. Resolution and frame rate are never changed. Audio that is already efficient (AAC, Opus…) is kept as is, without re encoding. Videos with several audio tracks or subtitles go to MKV with every track kept.
🖐️ You stay in control: applying only fills the usual settings, every setting stays visible and editable, and nothing starts until you click CONVERT. Ignore the card, or turn the automatic suggestions off in Destination ("Suggest compression optimizations automatically").
✦ New Optimized compression preset: a gold button in Output format that analyses the selected video and applies settings adapted to it. If the video is already efficient, it explains why and lets you "Apply anyway".
📏 Honest size and time estimates
📐 Indicative estimate: an instant size range based on the source and the target encoder, clearly marked as indicative.
🔬 Measure size and time: FFmpeg Forge encodes three short samples (at 20 %, 50 % and 80 % of the video) with the recommended settings, at below normal priority, and extrapolates the real size and the real speed of your computer. In our tests, a 2 min 34 s H.264 video of 849 MB was measured at 132 to 161 MB and 10 to 14 minutes; the full conversion gave 146.7 MB in 12 min 50 s.
🧠 Time estimates also learn from your previous conversions (same codec, speed and resolution class). Without data, FFmpeg Forge says so instead of inventing a number. Estimates are always shown as estimates, never as guarantees.
⏱️ Remaining time during the export: the file list and the status bar now show the progress and the time left ("38 % · about 30 s left").
Cover / thumbnail
New "Cover / thumbnail" option in Video settings: a frame of the converted video at the time you choose, or your own picture. It is added after the conversion, without re encoding the video. MP4 and M4V get cover art (shown by Windows Explorer and most players); MKV gets a cover.jpg attachment. Other formats (MOV, WebM, AVI…) do not support it reliably and are skipped with a message.
NVIDIA RTX Video in Video and Image
New "NVIDIA RTX Video" list (Off, ×2, ×3, ×4) in Video settings and Image settings, using NVIDIA RTX Video Super Resolution. For videos it is linked to the NVIDIA RTX AI section. For pictures it is new: a single picture (or a frame of a video) goes through the built in NVIDIA driver engine (ForgeRTX). Needs an NVIDIA RTX GPU with "RTX Video Enhancement" enabled in the NVIDIA App.
PNG/JPG SEQUENCE
The image sequence export in Video settings is now a simple "PNG/JPG SEQUENCE" option: tick PNG or JPG to export your video as numbered pictures, nothing ticked for a normal video file.
Unchanged and preserved
The live before / after comparison, the progress bar, cancelling, every preset, codec and format, the AI features (Denoiser IA, Python / ONNX AI, VisionNext, SeedVR2) and your saved settings work exactly as before.
Enjoying FFmpegForge? If you find it useful, please consider stars the project on GitHub! It helps more people discover this free software and supports its continued development.
[NEW] FFmpeg Forge 2.0
✨ FFmpeg Forge 2.0.0
🎨 A brand new design: royal blue and gold premium dark theme, gold logo, rounded cards, gradient buttons and a blue to gold progress bar. The layout stays the one you know from 1.1.5: header on top, your files in the middle and every setting in one column on the right.
🧠 Python / ONNX AI: a new section where you import your own AI models and run them on every frame of your videos and images.
ONNX models (.onnx) run with ONNX Runtime GPU (CUDA) or on the CPU, NCHW or NHWC. Models with a fixed input size are tiled automatically.
PyTorch models (.pth .pt .ckpt .safetensors .bin) are loaded by spandrel: ESRGAN, Real ESRGAN, SwinIR, HAT, DAT, SPAN, SCUNet, NAFNet, Restormer, OmniSR and many more.
Python scripts (.py) with load(device) or process(img). The "Python template" button writes a ready to edit example.
The model scale (×1, ×2, ×4…) is detected automatically. Options: keep the original size, GPU or CPU, fp16 or fp32, tile size. It runs before Scale, RTX, SeedVR2 and VisionNext and keeps your format, codec, audio, trim and filters. "Install AI runtime" installs PyTorch CUDA, ONNX Runtime GPU and spandrel once.
🖼️ PNG/JPG SEQUENCE: in Video settings, tick PNG (lossless, 16 bit when the render is 10 bit) or JPG (quality 2, full 4:4:4 colour) to export your video as pictures. Every frame of the final render is saved as name_000001.png… in a name_PNG (or name_JPG) folder next to the video, with every video option and AI effect applied: Denoiser IA, Python / ONNX AI, Scale, OptiScaler, NVIDIA RTX, SeedVR2 and Upscale VisionNext.
🧹 Denoiser IA models: choose the AI in a list. SCUNet (built in, real noise and film grain), Nomos8k HAT L (finest detail, very heavy), NomosUni DAT (strong denoising, natural look) or PurePhoto SPAN (light clean up, very fast). The ×4 models are brought back to the source size, so the picture is cleaned without changing its resolution.
🚀 Faster Denoiser IA: the whole SCUNet network is replayed as a CUDA Graph (one call per batch instead of about 8,500 small GPU launches), with fused window attention and up to 8 frames per pass.
🔭 Faster Upscale VisionNext: 2 frames per GPU pass on 20 GB cards and above, the bicubic pre resize runs on the CPU in parallel and the colour fix stays on the GPU. The pixels are exactly the same as before. Upscale factor ×1, ×2, ×3, ×4 and ×8, tile size 256 or 512.
🛡️ The live before / after preview can no longer make an export fail: if its pictures cannot be written, only the preview stops and the export goes on.
✅ Everything from 1.1.5 is included: VisionNext built in (GPU or CPU, videos encoded directly), SeedVR2, OptiScaler, Scale, NVIDIA RTX AI, CRF NVENC, DSF and DFF in every channel layout, Dolby Atmos and DTS:X layouts, Spatial Audio, filters and all formats.
⚙️ Requirements for the AI features
An NVIDIA GPU (CUDA) with a recent driver is recommended. Without one, the AI runs on the CPU, which is much slower.
About 3 GB of disk space for the AI runtime (Denoiser IA and Python / ONNX AI), and about 10 GB for VisionNext built in (24 GB of VRAM recommended).
🎒 FFmpegForge-2.0.0-Portable.rar (about 206 MB): everything included, FFmpeg Forge 2.0.0 with FFmpeg 9.0.2 full and ForgeRTX. Extract the whole archive anywhere and run FFmpegForge.exe.
Enjoying FFmpegForge? If you find it useful, please consider stars the project on GitHub! It helps more people discover this free software and supports its continued development.
FFmpeg Forge 1.1.5 FINAL
✨ New in 1.1.5
🔭 Upscale VisionNext without Visual Enhancer: VisionNext is now built in. Its model is VOSR 2.0 (one step 1.4B generative super resolution, CVPR 2026, Apache 2.0). Click "Install VisionNext model…" once and FFmpeg Forge installs the official VOSR code, the VOSR 2.0 weights with their VAE and DINOv2 encoder, Python and PyTorch CUDA in its own folder (VisionNext next to FFmpegForge.exe), then runs the model itself on your NVIDIA GPU.
🖥️ GPU or CPU: tick GPU (NVIDIA CUDA, fast) or CPU (works without an NVIDIA card, much slower). The choice appears as soon as Upscale VisionNext is ticked.
📈 Upscale factor ×1, ×2, ×3, ×4 and ×8, with tile size 256 or 512. The work grows with the output size: ×3 takes about 2 times longer than ×2, ×4 about 4 times and ×8 about 16 times. Very large outputs switch the VAE to tiles automatically so the GPU memory is not exceeded.
🎬 Videos are encoded directly, never picture by picture. With the built in engine, the video goes from the decoder to the model and straight into the final encoder: no picture files on disk and one single encode. With the Visual Enhancer engine, whose command line only reads pictures, FFmpeg Forge feeds it small batches of frames while the final encoder is already running, so there is still one single encode and no full frame sequence on disk.
⚡ Faster VisionNext: bf16 on the tensor cores with flash attention, all the tiles of a frame processed in one pass, and the same noise for every frame so there is no flicker between frames. Norms, tile blending and the colour fix stay in full precision. Triton is no longer needed (it does not exist on Windows), which fixes the "TritonMissing" error.
🧹 Denoiser IA: the Denoiser section is now a single checkbox. Tick it and FFmpeg Forge removes noise and film grain from videos and images with an AI network, for a clean and natural picture in the spirit of Neat Video. The AI is SCUNet (Swin Conv UNet by Kai Zhang et al., Apache 2.0), trained for real world noise: camera sensor noise, film grain and compression noise. A Neat Video style temporal stabilisation keeps the picture free of flicker.
🚀 Much faster Denoiser IA: the whole SCUNet network is captured once as a CUDA Graph and replayed in a single call per batch, instead of launching about 8,500 small GPU operations one by one. The window attention is fused into one GPU kernel, up to 8 frames are processed per pass, in fp16, with decoding, AI work and encoding running at the same time. If the GPU memory is full, the work size shrinks automatically, and the normal mode is used if CUDA Graphs are not available.
👁️ Live preview: the before / after window closes by itself when the export is finished, and it can no longer make an export fail. If its pictures cannot be written (TEMP folder cleaned by Windows or an antivirus, file locked), only the preview stops and the export goes on.
🛠️ Fixes: clearer error messages for the built in VisionNext engine, a warning when the output is too large for H.264 NVENC (4096 px) or H.265 NVENC (8192 px), and picture sizes that are not a multiple of 8 now work with VisionNext.
✅ Everything from the 1.1.5 betas is included: live before / after preview, DSF and DFF in every channel layout (import and export), SeedVR2, OptiScaler, Scale, NVIDIA RTX AI, CRF NVENC, Spatial Music, filters and all formats.
⚙️ Requirements for the AI features
An NVIDIA GPU (CUDA) with a recent driver is recommended. Without one, the AI runs on the CPU, which is very slow.
About 3 GB of disk space for the Denoiser IA, and about 10 GB for VisionNext built in (24 GB of VRAM recommended, more for ×4 and ×8).
×3, ×4 and ×8 need the built in engine. For outputs larger than 4096 px choose libx264, libx265 or H.265.
📦 Downloads
🎒 FFmpegForge-1.1.5-Portable.rar (about 206 MB): everything included, FFmpeg Forge 1.1.5 with FFmpeg 9.0.2 full and ForgeRTX. Extract the whole archive anywhere and run FFmpegForge.exe. The AI Denoiser and VisionNext install next to it.
ℹ️ PyTorch, SCUNet and VOSR 2.0 are not included in the archives: the installers download them from their official sources (GitHub and Hugging Face).
Enjoying FFmpegForge? If you find it useful, please consider stars the project on GitHub! It helps more people discover this free software and supports its continued development.
FFmpeg Forge 1.1.5B (Bêta)
✨ New
👁️ Live before / after preview: while a video exports, a preview window opens and shows the original frame and the final rendered frame (Scale, FSR, OptiScaler, Denoiser, filters, RTX, SeedVR2, VisionNext…).
⏱️ The preview updates about once per second of video. The frames come from the same FFmpeg command as the export, so the video is decoded only once and the export is barely slower.
☑️ Option "Live before / after preview while exporting (beta)" in Video settings, ticked by default. If a file cannot be previewed, it is converted normally without the preview.
💿 DSD export: new audio formats DSF (Sony DSD Stream File) and DFF (Philips DSDIFF), and new audio codecs DSD64 2.8 MHz and DSD128 5.6 MHz.
🎚️ FFmpeg can read DSD but cannot write it, so FFmpeg Forge does it itself: FFmpeg prepares the audio (trim, channels, volume, filters, oversampling), then FFmpeg Forge's 1 bit 7th order delta sigma modulator writes the DSD stream. Measured THD+N (20 Hz to 20 kHz): about −113 dB in DSD64 and −124 dB in DSD128.
🔊 Mono, stereo, 3.0, quad, 5.0 and 5.1 (DSF and DFF hold up to 6 channels). 0 dBFS is written at 50 % modulation, the SACD reference level.
✅ Everything from 1.1.4 is still there: Upscale VisionNext, SeedVR2, OptiScaler, Denoiser, Scale, NVIDIA RTX AI, CRF NVENC, Spatial Music, filters and all formats.
📦 Downloads
🎒 FFmpegForge-1.1.5b-Portable.rar (about 204 MB): everything included, FFmpeg Forge 1.1.5b with FFmpeg 9.0.2 full and ForgeRTX. Extract the whole archive anywhere and run FFmpegForge.exe.
🪶 FFmpegForge-1.1.5b.rar (about 0.6 MB): normal edition for your own FFmpeg (C:\FFmpeg\bin or next to the .exe).
ℹ️ The live preview needs FFmpeg 4.4 or newer (included in the portable edition).
FFmpeg Forge 1.1.3.a
✨ New
🎚️ CRF for H.264 and H.265 NVIDIA NVENC: when the video codec is "H.264 GPU NVIDIA (NVENC)" or "H.265 GPU NVIDIA (NVENC)", a new "CRF NVENC" field appears in Video settings. Type any value from 1 to 51, like the CRF of x264 and x265.
1 gives the best quality and the biggest file, 51 the smallest file. Good values are 18 to 23 for H.264 and 20 to 26 for H.265.
0 keeps the automatic mode: the value comes from the Quality setting, as before.
A hint under the field tells you what the chosen value means (very high, high, medium or low quality).
The value is used as NVENC constant quality (FFmpeg rc vbr with cq), so the file size follows the picture, like a real CRF.
It works everywhere FFmpeg Forge encodes with NVENC: normal conversions, Scale (the screen quality option no longer changes your value) and SeedVR2.
A bitrate above 0 still takes priority over the CRF.
Your CRF is saved with your settings.
⚙️ Requirements
An NVIDIA GPU with NVENC (GeForce GTX 10 series or newer) and a recent driver.
📦 Download
🎒 FFmpegForge-1.1.3a-Portable.rar (about 204 MB): everything included, FFmpeg Forge 1.1.3a with FFmpeg 9.0.2 full and ForgeRTX. Extract the whole archive anywhere and run FFmpegForge.exe.
Important information: Future updates and versions will be available exclusively as portable versions. Thank you for your understanding.
FFmpeg Forge 1.1.3
✨ New
🌱 SeedVR2 for videos and images: ByteDance Seed's one step diffusion model. It removes blur, noise and compression artefacts and rebuilds fine detail. On videos, the result stays consistent over time.
🧠 Choose from 10 models:
3B: GGUF Q4_K_M, GGUF Q8_0, FP8 or FP16.
7B: GGUF Q4_K_M, FP8 mixed (recommended for 24 GB cards) or FP16.
7B Sharp: GGUF Q4_K_M, FP8 mixed or FP16.
The list shows the VRAM each model needs. Each model downloads automatically the first time you use it.
🎛️ Settings:
Output size: ×1.5, ×2, ×3, ×4, 720p, 1080p, 1440p, 4K or restore only.
Batch size and colour correction (LAB, wavelet, wavelet adaptive, HSV, AdaIN).
Input noise, latent noise and seed.
Chunking with temporal overlap for long videos.
VRAM savers: BlockSwap, VAE tiling and RAM offload.
Attention backend (SDPA, Flash Attention 2, SageAttention 2 or 3), torch.compile and CUDA GPU.
📥 "Install SeedVR2…" button: it downloads the official SeedVR2 command line (Apache 2.0), uv with Python 3.12 and PyTorch CUDA 12.8 (about 4 GB). It runs in a console window so you can follow the progress. An existing install works too, ComfyUI included: use "SeedVR2 folder…".
🔗 Keeps the rest of FFmpeg Forge: trim, rotation, Denoiser, filters, output format, codec and audio still apply. FFmpeg prepares the source, SeedVR2 restores it and FFmpeg encodes the final file.
💡 Clear hints for common problems: out of VRAM, incomplete Python environment, CUDA not available.
🔧 Also in 1.1.3
The experimental DLSS and DLSS 5 options added after 1.1.2 have been removed.
The startup fix of the portable edition is kept, along with the error log (FFmpegForge_error.log).
⚙️ Requirements for SeedVR2
An NVIDIA GPU with CUDA and a recent driver. For the 7B models, 24 GB of VRAM is recommended; use 3B, GGUF or BlockSwap on smaller cards.
About 4 GB for the installation, plus 2 to 16 GB per model.
It is slow: a 7B model needs several minutes per minute of video.
📦 Downloads
🎒 FFmpegForge-1.1.3-Portable.rar (about 204 MB): everything included, FFmpeg Forge 1.1.3 with FFmpeg 9.0.2 full and ForgeRTX. Extract the whole archive anywhere and run FFmpegForge.exe. SeedVR2 installs into the SeedVR2 folder next to it.
🪶 FFmpegForge-1.1.3.rar (about 0.6 MB): normal edition for your own FFmpeg (C:\FFmpeg\bin or next to the .exe). SeedVR2 installs into %LOCALAPPDATA%\FFmpegForge\SeedVR2.
ℹ️ SeedVR2, Python, PyTorch and the models are not included: the "Install SeedVR2…" button downloads them from their official sources.
🙏 Credits
SeedVR2 by ByteDance Seed · ComfyUI SeedVR2 VideoUpscaler by numz and AInVFX (Apache 2.0) · FFmpeg · uv · PyTorch