Recherche avancée

Sur d’autres sites (9009)

• How to improve Desktop capture performance and quality with ffmpeg [closed]

  6 novembre 2024, par Francesco Bramato

  I'm developing a game capture feature from my Electron app. I'm working on this since a while and tried a lot of different parameters combinations, now i'm running out of ideas :)

  


  I've read tons of ffmpeg documentation, SO posts, other sites, but i'm not really a ffmpeg expert or video editing pro.

  


  This is how it works now :

  


  The app spawn an ffmpeg command based on user's settings :

  


  

  • Output format (mp4, mkv, avi)

  


  • Framerate (12, 24, 30, 60)

  


  • Codec (X264, NVidia NVENC, AMD AMF)

  


  • Bitrate (from 1000 to 10000kpbs)

  


  • Presets (for X264)

  


  • Audio output (a dshow device like StereoMix or VB-Cable) and Audio input (a dshow device like the Microphone)

  


  • Final Resolution (720p, 1080p, 2K, Original Size)

  


  


  The command executed, as far, is :

  


  ffmpeg.exe -nostats -hide_banner -hwaccel cuda -hwaccel_output_format cuda -f gdigrab -draw_mouse 0 -framerate 60 -offset_x 0 -offset_y 0 -video_size 2560x1440 -i desktop -f dshow -i audio=@device_cm_{33D9A762-90C8-11D0-BD43-00A0C911CE86}\wave_{D61FA53D-FA37-4BE7-BE2F-4005F94790BB} -ar 44100 -colorspace bt709 -color_trc bt709 -color_primaries bt709 -c:v h264_nvenc -b:v 6000k -preset slow -rc cbr -profile:v high -g 60 -acodec aac -maxrate 6000k -bufsize 12000k -pix_fmt yuv420p -f mpegts -


  


  one of the settings is the recording mode : full game session or replay buffer.
In case of full game session, the output is a file, for replay buffer is stdout.

  


  The output format is mpegts because, as far i have read in a lot of places, the video stream can be cut in any moment.

  


  Replays are cutted with different past and future duration based on game events.

  


  In full game session, the replays are cutted directly from the mpegts.

  


  In replay buffer mode, the ffmpeg stdout is redirect to the app that record the buffer (1 or 2 minutes), when the replay must be created, the app saves on the disk the buffer section according to past and future duration and with another ffmpeg command, copy it to a mp4 or mkv final file.

  


  Generally speaking, this works reliably.

  


  There are few issues :

  


  

  • nonetheless i ask ffmpeg to capture at 60fps, the final result is at 30fps (using -r 60 will speed up the final result)

  


  • some user has reported FPS drops in-game, specially when using NVidia NVENC (and having a NVIDIA GPU), using X264 seems save some FPS

  


  • colors are strange compared to original, what i see on screen, they seem washed out - i could have solved this using -colorspace bt709 -color_trc bt709 -color_primaries bt709 but don't know if is the right choice

  


  • NVIDIA NVenc with any other preset that is not slow creates videos terribly laggy

  


  


  here two examples, 60 FPS, NVIDIA NVENC (slow, 6000kbs, MP4

  


  Recorded by my app : https://www.youtube.com/watch?v=Msm62IwHdlk

  


  Recorded by OB with nearly same settings : https://youtu.be/WuHoLh26W7E

  


  Hope someone can help me

  


  Thanks !

  


• How can I decode MegaRace (93/94) .HNM videos ? — FFmpeg fails and legacy tools are missing [closed]

  29 juillet, par L P

  Goal

  


  

  • I want to convert the original full‑motion videos shipped with MegaRace (DOS, 1993/94) from Cryo’s proprietary HNM format into a modern, playable container (e.g. lossless AVI/PNG sequence).

  


  • Capturing the screen while the game runs is not an option : the target is the pure background animation (no sprites, no overlays, no DOSBox scaling filters, no subtitles).

  


  


  What I already have

  


  

  • Extracted all *.HNM files out of MEGARACE.DAT with Game Extractor.
    
(Directory listing : 100+ files from AZUR.HNM to VWBT5.HNM.)

  


  • Verified that they are indeed Cryo HNM : the wiki pages [1] and [2] say MegaRace uses an early variant (v1/v2) written by Pascal Urro (might also used in 92's Dune, 95's Lost Eden etc.).

  


  • Tried FFmpeg (from latest to earliest) with ffplay -f hnm -i EXPLO.HNM, but every file aborts with invalid resolution and invalid data found when processing input. Same for ffprobe and ffmpeg -i.

  


  • Searched for legacy tools :

    

    • hnmdec / hnm2avi — no longer included in modern ScummVM‑Tools ; seems to have been present at some point.

    


    • extract_cryo — source vanished.

    


    • Game Media Formats DirectShow Filters, by Valery Anisimovsky could, in theory, load the vendor codecs (CM6_*.dll) and expose them to DirectShow, but I haven't found the necessary Cryo DLLs shipped with MegaRace.

    


    


  


  • Looked at the DOSBox “Video Codec” folder – it only contains ZMBV (zmbv.dll, zmbv.inf), which is unrelated (that’s the codec DOSBox uses for screen‑capture).

  


  


  Environment

  


  

  • macOS Sequoia 15.5 on Apple Silicon (but I can boot Windows 11 ARM or use x86_64 VMs if the solution requires them).

  


  • No need for audio ; video stream alone is fine.

  


  


  Why this matters

  


  

  • Archiving early FMV assets is part of game preservation. MegaRace’s CG backgrounds are unique and currently inaccessible without running the whole DOS game.

  


  


  What I’m looking for

  


  

  • Specifications or source code for the HNM v1/v2 codec.

  


  • A working decoder (CLI or library) for any platform so I can batch‑convert all files.

  


  • Pointers on how to patch FFmpeg’s current libavcodec/hnm4video.c to support the older variant.

  


  • Alternatively, hints on extracting the player’s built‑in decoder from MEGARACE.EXE (or ScummVM’s internal implementation, if any) and wrapping it.

  


  • Any lead, paper, or old DOS utility is appreciated !

  


  


• Empty audio backends for torchaudio list audio backends, with ffmpeg installed on system libraries [closed]

  28 mars, par Alberto Agudo Dominguez

  I installed torchaudio 2.5.1 and a system install of ffmpeg on Windows and get :

  


  PS C:\Users\> ffmpeg -version
ffmpeg version 2025-01-05-git-19c95ecbff-essentials_build-www.gyan.dev Copyright (c) 2000-2025 the FFmpeg developers
built with gcc 14.2.0 (Rev1, Built by MSYS2 project)
configuration: --enable-gpl --enable-version3 --enable-static --disable-w32threads --disable-autodetect --enable-fontconfig --enable-iconv --enable-gnutls --enable-libxml2 --enable-gmp --enable-bzlib --enable-lzma --enable-zlib --enable-libsrt --enable-libssh --enable-libzmq --enable-avisynth --enable-sdl2 --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxvid --enable-libaom --enable-libopenjpeg --enable-libvpx --enable-mediafoundation --enable-libass --enable-libfreetype --enable-libfribidi --enable-libharfbuzz --enable-libvidstab --enable-libvmaf --enable-libzimg --enable-amf --enable-cuda-llvm --enable-cuvid --enable-dxva2 --enable-d3d11va --enable-d3d12va --enable-ffnvcodec --enable-libvpl --enable-nvdec --enable-nvenc --enable-vaapi --enable-libgme --enable-libopenmpt --enable-libopencore-amrwb --enable-libmp3lame --enable-libtheora --enable-libvo-amrwbenc --enable-libgsm --enable-libopencore-amrnb --enable-libopus --enable-libspeex --enable-libvorbis --enable-librubberband
libavutil      59. 54.101 / 59. 54.101
libavcodec     61. 31.100 / 61. 31.100
libavfilter    10.  6.101 / 10.  6.101
libswresample   5.  4.100 /  5.  4.100
libpostproc    58.  4.100 / 58.  4.100

PS C:\Users\> python
Python 3.12.5 (tags/v3.12.5:ff3bc82, Aug  6 2024, 20:45:27) [MSC v.1940 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import torchaudio
>>> torchaudio.list_audio_backends()
[]


  


  Hence ffmpeg is added to the path and recognized by the console, but not by torchaudio.