#VMAF
Socialync now runs its own video encode service and it scores 95.8 on VMAF against the originals. One clean 1080p for all eight platforms, clean exports untouched. The file we hand TikTok looks like the file you uploaded. Built it solo.
September 18, 2026 at 3:00 PM
How to use libvmaf_cuda on Windows: an easy-to-follow guide (WSL2 + Docker + NVIDIA)
**TL;DR:** Running GPU-accelerated VMAF on Windows is surprisingly painful. After spending hours fighting broken builds and undocumented errors, the only reliable path I found is **WSL2 + Docker Desktop + NVIDIA Container Toolkit + a CUDA-enabled FFmpeg**. This guide walks you through the whole setup using the easyVmaf project, so you can skip the trial-and-error I went through. This guide uses **easyVmaf** , a project that ships a `Dockerfile.cuda` specifically built for this purpose. We'll run it inside WSL2, with Docker Desktop and the NVIDIA Container Toolkit handling the GPU passthrough. **The result:** VMAF running at ~15x real-time speed on an RTX 3060 Mobile. A 46-minute video gets analyzed in about 3 minutes. On CPU, the same task would take like an hour. > ⚠️ **AMD and Intel GPUs will not work with this guide.** `libvmaf_cuda` is NVIDIA-exclusive. ## Table of contents * The Problem * Prerequisites * Part 1 — Setting up the environment * 1.1 Install WSL2 * 1.2 Accessing Windows drives from WSL * 1.3 Verify GPU access from WSL * 1.4 Install Docker Desktop * 1.5 Install the NVIDIA Container Toolkit * Part 2 — Building the easyVmaf image * 2.1 Clone the repository * 2.2 Critical fix: pin nv-codec-headers * 2.3 Build the image * 2.4 Verify the libvmaf_cuda filter * Part 3 — Running VMAF with GPU acceleration * 3.1 capabilities=video * 3.2 Full analysis command * Troubleshooting * Conclusion * References * Let's Connect ## The Problem If you've ever tried to calculate VMAF on Windows, you already know that: * The standard `libvmaf` filter works, but it runs on the **CPU** (slow). * `libvmaf_cuda` **does not exist in any prebuilt Windows binary** (not in gyan.dev, not in BtbN, not anywhere _that I know_). * Most information online is scattered, outdated, or simply missing. ## Prerequisites Before starting, make sure you have: Component | Minimum requirement ---|--- **Windows** | Windows 10 (version 2004+) or Windows 11 **NVIDIA GPU** | Any CUDA-capable GPU (I'm using an RTX 3060 Mobile) **NVIDIA drivers** | Version 525+ (for CUDA 12.x) — download here **Disk space** | ~30 GB free (WSL + Docker + images) **RAM** | 16 GB recommended (WSL2 is memory-hungry) > ⚠️ **Do not install NVIDIA drivers inside WSL.** WSL automatically uses the drivers from Windows. Installing Linux drivers inside WSL will break GPU passthrough. ## Part 1 — Setting up the environment This part covers everything needed to get a working Linux + Docker + GPU stack: WSL2, Docker Desktop, and the NVIDIA Container Toolkit. ### 1.1 Install WSL2 Open **PowerShell as Administrator** and run wsl --install This command will: * download and install WSL2 kernel * install Ubuntu distro by default **Restart Windows** when prompted. After the restart, open PowerShell again and verify: wsl --list --verbose Expected output: NAME STATE VERSION * Ubuntu Running 2 Make sure `VERSION` is `2`. If it says `1`, upgrade with: wsl --set-version Ubuntu 2 wsl --set-default-version 2 #### Ubuntu update Open your Ubuntu terminal either by typing in PowerShell ubuntu or wsl Once in Ubuntu terminal enter: sudo apt update && sudo apt upgrade -y ### 1.2 Accessing Windows drives from WSL If you're new to Linux, one of the first things to understand is that WSL doesn't use `C:\`, `D:\`, etc. Instead, it mounts every Windows drive under `/mnt/`. Windows path | WSL path ---|--- `C:\Users\YourName\Videos` | `/mnt/c/Users/YourName/Videos` `D:\Movies` | `/mnt/d/Movies` `E:\Backups\2026` | `/mnt/e/Backups/2026` ### 1.3 Verify GPU access from WSL if you already have installed NVIDIA drivers on Windows, run: nvidia-smi You should see the `nvidia-smi` table with your GPU listed. If it doesn't work, your Windows drivers are outdated or WSL isn't configured properly. Wed Sep 16 09:30:35 2026 +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 615.71.08 KMD Version: 616.92 CUDA UMD Version: 13.4 | +-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce RTX 3060 ... On | 00000000:01:00.0 On | N/A | | N/A 55C P8 14W / 115W | 1085MiB / 6144MiB | 6% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | No running processes found | +-----------------------------------------------------------------------------------------+ ### 1.4 Install Docker Desktop Download Docker Desktop from: https://www.docker.com/products/docker-desktop/ During installation, make sure to check **"Use WSL 2 instead of Hyper-V"**. #### Configure WSL2 integration Once installed, open Docker Desktop and go to Settings **General tab:** * Check **Use the WSL 2 based engine** * Click **Apply & Restart** **Resources → WSL Integration:** * Check **Enable integration with my default WSL distro** * Turn on the Ubuntu toggle **Resources → Advanced (optional):** * Uncheck **"Enable Resource Saver"** > If "Resource Saver" is enabled, Docker will suspend WSL2 after inactivity, and you'll have to restart it manually. #### Verify Docker works in WSL In your Ubuntu terminal: docker --version Expected output: Docker version 27.3.1, build ce12230 > ⚠️ If you get `var/run/docker.sock: connect: permission denied.` jump to the Troubleshooting section. ### 1.5 Install the NVIDIA Container Toolkit This is the component that allows Docker containers to access the GPU. #### Install In your WSL Ubuntu terminal: curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | \ sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg && \ curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \ sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \ sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list sudo apt update sudo apt install -y nvidia-container-toolkit #### Configure the Docker runtime sudo nvidia-ctk runtime configure --runtime=docker Expected output: INFO[0000] Config file does not exist; using empty config INFO[0000] Wrote updated config to /etc/docker/daemon.json INFO[0000] It is recommended that docker daemon be restarted. #### Restart Docker Restart **Docker Desktop** from Windows (either via the restart icon or by quitting and reopening it). #### Verify Docker can see the GPU Once Docker Desktop is running again, execute in your WSL terminal: docker run --rm --gpus all nvidia/cuda:12.3.2-base-ubuntu22.04 nvidia-smi You should see the `nvidia-smi` table with your GPU listed. +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 615.71.08 KMD Version: 616.92 CUDA UMD Version: 13.4 | +-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce RTX 3060 ... On | 00000000:01:00.0 On | N/A | | N/A 55C P8 14W / 115W | 1085MiB / 6144MiB | 6% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | No running processes found | +-----------------------------------------------------------------------------------------+ ## Part 2 — Building the easyVmaf image We'll use the easyVmaf project, which ships a `Dockerfile.cuda` configured for GPU-accelerated VMAF. ### 2.1 Clone the repository cd ~ git clone --depth 1 https://github.com/gdavila/easyVmaf.git cd easyVmaf ### 2.2 Critical fix: pin nv-codec-headers > 💡 **This step is undocumented anywhere else.** I figured it out after spending hours debugging the compilation error with the help of AI. Skip it and your build will fail. The original `Dockerfile.cuda` clones the **latest** version of `nv-codec-headers`, which is **incompatible with FFmpeg 8.1**. You'll get this error during the build: libavcodec/nvenc.c:2529:42: error: 'NV_ENC_CLOCK_TIMESTAMP_SET' has no member named 'countingType'; did you mean 'countingTypeLSB'? In recent versions, NVIDIA renamed `countingType` to `countingTypeLSB/countingTypeMSB`, but FFmpeg 8.1 still uses `countingType`. **The fix:** Pin `nv-codec-headers` to version `n12.1.14.0`, which still uses `countingType` **and** already includes the modern CUDA functions (`cuStreamCreateWithPriority`, `cuMemHostAlloc`, etc.) that `libvmaf_cuda` needs. Edit the Dockerfile: nano Dockerfile.cuda * Press `CTRL+W`, type `nv-codec-headers`, press `ENTER`. You'll land on this line: RUN git clone --depth 1 https://git.videolan.org/git/ffmpeg/nv-codec-headers.git && \ cd nv-codec-headers && \ make install Change it by adding `n12.1.14.0` before --depth: RUN git clone --branch n12.1.14.0 --depth 1 https://git.videolan.org/git/ffmpeg/nv-codec-headers.git && \ cd nv-codec-headers && \ make install Save with `CTRL+X`, then `Y`, then `ENTER`. ### 2.3 Build the image docker build -f Dockerfile.cuda -t easyvmaf:cuda . ### 2.4 Verify the `libvmaf_cuda` filter > ⚠️ The `easyvmaf:cuda` image defines `easyVmaf` (its own CLI) as the `ENTRYPOINT`. To run raw `ffmpeg`, you must override the entrypoint. > docker run --rm --gpus all --entrypoint ffmpeg easyvmaf:cuda -filters | grep -E "libvmaf|scale_cuda" Expected output: .. libvmaf VV->V Calculate the VMAF between two video streams. .. libvmaf_cuda VV->V Calculate the VMAF between two video streams. .. scale_cuda V->V GPU accelerated video resizer **If you see`libvmaf_cuda`, you're done with the setup.** ## Part 3 — Running VMAF with GPU acceleration ### 3.1 capabilities=video By default, `--gpus all` alone only grants the compute and utility capabilities. It does not mount the video decode/encode libraries `libnvcuvid.so.1` (NVDEC) and `libnvidia-encode.so.1` (NVENC). Since we tell FFmpeg to decode with `-hwaccel cuda`, it needs those libraries. Without them, FFmpeg fails with: Cannot load libnvcuvid.so.1 Failed loading nvcuvid. Failed setup for format cuda: hwaccel initialisation returned error. **The fix:** explicitly request the `video` capability: --gpus all,capabilities=video Now Docker mounts `libcuda.so.1` (CUDA compute), `libnvcuvid.so.1` (NVDEC), and `libnvidia-encode.so.1` (NVENC). FFmpeg can then decode, filter, and analyze entirely on the GPU. ### 3.2 Full analysis command docker run --gpus all,capabilities=video --rm --entrypoint ffmpeg \ -v "/path/to/your/videos":/videos \ easyvmaf:cuda \ -hwaccel cuda -hwaccel_output_format cuda \ -i "/videos/distorted.mkv" \ -hwaccel cuda -hwaccel_output_format cuda \ -i "/videos/reference.mkv" \ -filter_complex "[0:v]scale_cuda=format=yuv420p[dis];[1:v]scale_cuda=format=yuv420p[ref];[dis][ref]libvmaf_cuda=log_fmt=json:log_path=/videos/vmaf_full.json" \ -f null - #### Real-world result Here's the output from a test on a 46-minute video file: [Parsed_libvmaf_cuda_2 @ 0x760b44004f80] VMAF score: 93.558906 speed=14.9x elapsed=0:03:05.05 [out#0/null @ 0x5ef225e22140] video:27438KiB audio:2072848KiB subtitle:0KiB frame=66265 fps=350 q=-0.0 Lsize=N/A time=00:46:03.79 bitrate=N/A speed=14.6x elapsed=0:03:09.43 **3 minutes for a 46-minute video at ~15x real-time speed.** That's the whole point of using `libvmaf_cuda`. ## Troubleshooting ### `var/run/docker.sock: connect: permission denied` docker: permission denied while trying to connect to the Docker daemon socket at unix:///var/run/docker.sock: Head "http://%2Fvar%2Frun%2Fdocker.sock/_ping": dial unix /var/run/docker.sock: connect: permission denied. Your user doesn't belong to the `docker` group, which owns `/var/run/docker.sock`. fix: enter this command in your WSL Ubuntu terminal sudo usermod -aG docker $USER This adds your user to the `docker` group. Then **close and reopen WSL** (group changes only apply to new sessions), and verify: groups Expected output: youruser adm cdrom sudo dip plugdev users docker ### `Cannot load libnvcuvid.so.1` Missing `,capabilities=video` in the `--gpus` flag. See section 3.1. ### `NV_ENC_CLOCK_TIMESTAMP_SET has no member named 'countingType'` You didn't pin `nv-codec-headers` to `n12.1.14.0`. See section 2.2. ### WSL2 hangs after inactivity Edit `C:\Users\YOUR_USER\.wslconfig`: [wsl2] vmIdleTimeout=-1 Then run `wsl --shutdown` in PowerShell. Also disable "Resource Saver" in Docker Desktop (see section 1.3). ## Conclusion Setting up `libvmaf_cuda` on Windows was a long journey. At the start, I couldn't find much information about it — most guides either stop at "use `libvmaf` on CPU" or assume you're on Linux. Even though this isn't 100% native to Windows (it runs through WSL2 + Docker), it's a solid alternative that is absolutely worth the effort. Once it's working, you get VMAF analysis at **15x real-time speed** , which completely changes what's practical for video quality workflows. ## References * easyVmaf — the project powering this guide * NVIDIA Container Toolkit docs * WSL2 GPU support * Netflix VMAF ## Let's Connect If you found this post useful and you're looking for a Full Stack Developer or a technical writer, feel free to reach out! * GitHub * **Telegram** : @rtaglia Thanks for reading! 🙌
dev.to
September 16, 2026 at 7:40 PM
🐧 **ab-av1 – AV1 encoding with fast quality sampling**

ab-av1 is a video encoding tool that samples VMAF or XPSNR, searches for suitable CRF values, predicts output size, and drives FFmpeg encodes. The post ab-av1 – AV1 encoding with fast quality sampl...

📰 Source: LinuxLinks
🔗 Link […]
Original post on igeek.gamer-geek-news.com
igeek.gamer-geek-news.com
September 16, 2026 at 7:19 PM
[VCEEnc 9.17]
--vpp-kfmのmode=24/60のRFF対応を実施したほか、VMAFによる画質評価を追加しました。
rigaya34589.blog.fc2.com/blog-entry-2...
September 11, 2026 at 1:29 PM
[svtAV1guiEx 2.11]
AviUtl2でプロジェクト単位の出力設定の保存・復元に対応したほか、svt-av1 4.2.0に合わせ、いくつかのオプションの設定欄を追加・更新しました。
rigaya34589.blog.fc2.com/blog-entry-2...
svtAV1guiEx 2.11
AviUtl2でプロジェクト単位の出力設定の保存・復元に対応。svt-av1 4.2.0に合わせ、設定欄を追加。  --tune 4: MS-SSIM / 5: VMAF を追加  --enable-dlf 2=accurate を追加  --fast-decode 2を追加  CQPを0.25刻み・上限70に。  下記を新規追加   --enable-...
rigaya34589.blog.fc2.com
July 25, 2026 at 1:47 PM
VMAFスコア95を基準として目指そうとするとRTX5070tiを積んだWindowsマシンの方が早めに終わってくれるんですわ_(・ω・_
July 25, 2026 at 2:57 AM
SEQUENCE OF EVENTS: Three F-35Bs were dispatched from MCAS Miramar at 14:30:00 L to Los Angeles. While no flight plan or official briefing were filed, the MP and his wingmen told investigators that the CO VMAF 311 instructed them to perform low-altitude flight over ongoing public demonstrations.
July 17, 2026 at 11:20 PM
There were 59 fatalities and 72 serious injuries. The mishap pilot (MP), assigned to VMAF-311, ejected safely before impact. He sustained minor, non-life-threatening injuries. The MA was destroyed upon impact, with a total loss of the airframe valued at $196,500,000.
July 17, 2026 at 11:20 PM
On 18 November 2026, at approximately 15:30:47 local (L), the mishap aircraft (MA), an F-35B aircraft, tail number (T/N) 69-4201, crashed while performing a hover-in-flight maneuver in downtown Los Angeles. The MA was operated out of MCAS Miramar, by Marine Fighter Attack Squadron 311 (VMAF-311).
July 17, 2026 at 11:20 PM
[NVEnc 9.24]
--vpp-kfmの最適化を行い、高速化したほか、
RIFE v4.xフレーム補間フィルタ(--vpp-rife-ov)や--vpp-onnxへのsdr2hdrモデルの追加など、多数のフィルタの機能拡張や不具合修正を行いました。
rigaya34589.blog.fc2.com/blog-entry-2...
NVEnc 9.24
- 長時間処理時の--vpp-kfmを高速化。だいぶいろいろ最適化を重ねて、かなり速くなってきたと思う。1440x1080 MPEG2をRTX4080で551fps、RTX2070でも396fps。- --vpp-onnxにSDR to HDRモデル対応を追加。(--vpp-onnx)- 存在しないVMAF jsonモデル指定のエラーを明確化。(--vmaf)- OpenVINO RIFEフレーム...
rigaya34589.blog.fc2.com
July 11, 2026 at 11:50 AM
FRVQM changes that by turning VMAF into a real-time KPI for your live services.

🚀 What this means for your organization:

✅ Monitor viewer-perceived quality live, not after delivery
✅ Detect and fix issues before they impact audience experience

2/4
July 7, 2026 at 9:53 AM
🔴 What if VMAF (Netflix's famous video quality metric) could guide your live operations in real time?

When video quality drops, users notice immediately. But without the right tools, teams often lack clear, objective data to respond.

1/4
July 7, 2026 at 9:53 AM
VMAF v1: Good Is Not Good Enough
https://medium.com/netflix-techblog/vmaf-v1-good-is-not-good-enough-60d7e4244ea8?source=rss-c3aeaf49d8a4------2
medium.com
July 4, 2026 at 12:40 PM
[NVEnc 9.23]
vmaf計算にGPUを使用できるようにして高速化したほか、フィルタ関連の不具合の修正を行いました。
rigaya34589.blog.fc2.com/blog-entry-2...

1080pで 130.59 fps → 315.82 fps と高速化しています!
NVEnc 9.23
- libvmafのCUDA版に対応。(--vmaf)いただいた要望の反映。1080pの--vmaf比較だけど、かなり速くなってる。- --vpp-finedehalo の高精度化。(--vpp-finedehalo)ご指摘いただいた問題の修正。- --vpp-finedehalo をインタレ非対応と明示。(--vpp-finedehalo)ご指摘いただいた問題の修正。- フィルタ関連の不具合の修...
rigaya34589.blog.fc2.com
July 4, 2026 at 12:07 PM
Start optimizing with real-time VMAF measurement and monitoring.

📩 Request a free evaluation:
acceptv.com/en/products_...

4/4
Full Reference Video Quality Monitor (includes VMAF) - FREE TRIAL
Full Reference Video Quality Monitor is a full-reference solution to measure video and audio quality and audio/video synchronization (lipsync) from live audio/video sources (IP streaming, HDMI, SDI, d...
acceptv.com
July 2, 2026 at 2:24 PM
It enables engineering teams to measure VMAF continuously on live signals, making it possible to validate encoding decisions as they happen.

🔬 With FRVQM, you can:

✅ Compare source vs encoded streams in real time using VMAF
✅ Fine-tune encoder settings based on live perceptual feedback

2/4
July 2, 2026 at 2:24 PM
🔴 Why is VMAF still mostly used offline whereas you can use it LIVE, in REAL TIME?

VMAF has become the gold standard for video quality evaluation, but too often it is limited to lab testing.

FRVQM removes that limitation.

1/4
July 2, 2026 at 2:24 PM
Netflix améliore son outil qui mesure la qualité des vidéos, pour vous offrir des images de meilleure qualité
Avez-vous déjà essayé de mesurer la qualité d'une vidéo ? Vous pouvez certes vérifier si des artefacts sont visibles, changer le câble HDMI ou le câble secteur pour espérer que la scène s'ouvre et même demander à votre femme si elle voit la différence. 1Ou tout simplement passer par un média physique, réputé meilleur. Mais vous pouvez aussi, comme Netflix, développer des outils pour une mesure à peu près factuelle, ou tout du moins automatisée. C'est la raison d'être du VMAF (Video Multimethod Assessment Fusion), une méthode open source qui passe de la v0 à la v1. Une représentation un peu simplifiée du but du VMAF : trouver le meilleur compromis entre débit et définition. Image Netflix. Le but du VMAF est de trouver le bon compromis entre la qualité perçue et la bande passante, le nerf de la guerre. Un débit plus faible permet une meilleure stabilité chez les utilisateurs, réduit les coûts chez Netflix et améliore globalement l'expérience… tant que l'image reste acceptable. Le problème de la v0 du VMAF, selon Netflix, c'est que la méthode tend à privilégier un flux en haute définition avec un débit bas à un flux basse définition avec un débit élevé, ce qui n'est pas toujours souhaitable. Dans le premier cas, on se retrouve plus fréquemment avec des artefacts de compression (les fameux macroblocs), dans le second, l'image est moins nette et le résultat est en partie dépendant du matériel. Un flux avec une définition plus faible peut être acceptable sur un petit téléviseur ou un smartphone, ou même sur un téléviseur avec un bon upscale, mais à l'inverse peut devenir gênant sur un grand écran. Un usage très concret : optimiser ce qu'on regarde Les outils peuvent sembler inutiles, et vous pouvez vous demander pourquoi Netflix voudrait noter la qualité des vidéos. La raison est très concrète : trouver le meilleur compromis pour la diffusion. Si encoder avec un débit élevé comme sur les disques optiques est une solution qui peut sembler valable, elle ne l'est pas : tout le monde n'a pas une connexion à plusieurs dizaines de mégabits/s pour regarder un film en 1080p. Netflix (et les autres services) utilisent donc le VMAF avec différentes versions d'un contenu (code, débit, définition, etc.) pour essayer de trouver le sweet spot, c'est-à-dire le compromis entre la bande passante, la définition et la qualité perçue. Le but n'est pas d'obtenir un résultat totalement parfait (c'est le créneau des disques et de certaines offres spécifiques) mais bien quelque chose de satisfaisant dans la majorité des cas. Une des raisons est que la base du VMAF est une personne qui regarde un contenu en 1080p dans une pièce standard à une distance qui est de l'ordre de trois fois la hauteur de l'écran (un téléviseur de 55 pouces à 2 mètres pour se donner une idée). Dans la v1, ce point a été modifié, avec une adaptation plus fine au contenu, parce que (par exemple) les artefacts de compression sont moins visibles sur un smartphone à cause de la taille réduite des écrans. L'analyse ne prend évidemment pas en compte ceux qui scrutent les pixels en cherchant celui qui n'a pas la bonne couleur. Image Netflix. VMAF v1 prend aussi en compte le banding, ce problème qui arrive parfois sur des dégradés qui sont affichés avec des bandes séparées bien visibles et les artefacts liés à la couleur (chroma), qui touchent parfois le rouge. Avec VMAF v1, des outils pour simuler l'absence d'améliorations sont aussi présents par défaut. Si beaucoup de téléviseurs disposent de modules d'upscale et d'améliorations de l'image performantes, ce n'est pas systématique, et les outils sont donc plus efficace en partant du principe que les améliorations en question sont absentes. Que préférez-vous ? (Image Netflix. Dans la pratique, le nouveau modèle prend donc en compte un utilisateur qui a un téléviseur 1080p, un smartphone (avec une distance de cinq fois la hauteur ici) et un écran 4K. Le calcul se fait ici sur une distance identique au 1080p (trois fois la hauteur) mais aussi avec une valeur plus faible (1,5x), probablement pour des raisons pragmatiques : la taille des salons n'augmente pas en parallèle de la diagonale des téléviseurs. Selon Netflix, VMAF v1 offre un meilleur résultat que VMAF v0 et (surtout) le fait plus rapidement. Il y a toujours des optimisations et des changements à prévoir pour prendre en compte certaines évolutions modernes, mais c'est une avancée intéressante pour une tâche souvent invisible mais importante. --- * Soyons clairs : c'est de l'ironie sur certains adeptes de l'audio, pas du sexisme au premier degré.  ↩︎
dlvr.it
July 1, 2026 at 4:52 PM
VMAF v1: Good Is Not Good Enough
https://netflixtechblog.com/vmaf-v1-good-is-not-good-enough-60d7e4244ea8?source=rss-c3aeaf49d8a4------2
netflixtechblog.com
July 1, 2026 at 12:40 AM
Happy to say VMAF v1 models will actually give dogshit quality video a straight up 0 instead of being a centrist and giving every compression artefact feverdream a 50
June 26, 2026 at 10:46 PM
VMAF v1: Good Is Not Good Enough | Stories by Netflix Technology Blog on Medium
VMAF v1: Good Is Not Good Enough
medium.com
June 22, 2026 at 10:14 PM