Skip to content

[Bug]: bad quality of desktop screen #1465

New issue
New issue
Open
Open
[Bug]: bad quality of desktop screen#1465
Labels
bugSomething isn't working

Description

@dealenx
dealenx
opened on Nov 2, 2026

Search existing issues

  • I have searched the existing issues

Describe the bug

Image

Expected behavior

In editor bad quality, after render also bad quality on desktop screen

To Reproduce

No response

Screenshots

No response

OS

None

OS Version

CachyOS x86_64 (Linux 7.2.9-1-cachyos) + niri 26.04 (Wayland)

Other OS

No response

Browser

None

Browser Version

No response

Other Browser

No response

Device Type

None

Other Device

No response

Additional context

                             dealenx@cachyos-asus-tuf-x8664
       ...........           ------------------------------
      /,,,,...../              OS  CachyOS x86_64
     /,,,..,,../   ()           Kernel  Linux 7.2.9-1-cachyos
    /,,,,../                   Uptime  8 hours
   /,,,.../     /'\            Packages  1301 (pacman)
  /,...../      \,/            Shell  fish 4.9.3
  \,,,,,,\\             _      WM  niri 26.04 (Wayland)
   \,,....\\           / \      Term  alacritty 0.17.0
    \...,..\\          \_/     󰻠 CPU  AMD Ryzen 5 3550H (8) @ 3.70 GHz
     \..,,,,,,,,,,,,/          󰢮 GPU  NVIDIA GeForce GTX 1650 Mobile / Max-Q [Discrete]
      \,,........,,/           󰍛 Memory  14.46 GiB / 29.30 GiB (49%)
                               󰓡 Swap  2.52 GiB / 29.30 GiB (9%)
                               󰋊 Disk (/)  329.59 GiB / 472.94 GiB (70%) - btrfs

Activity

  1. dealenx
    added
    bugSomething isn't working
    on Nov 2, 2026

  2. EtienneLescot commented on Nov 2, 2026

    @EtienneLescot
    EtienneLescot
    on Nov 2, 2026
    Member

    Thanks for the report!

    This looks related to #1450 (also CachyOS on Wayland). To narrow it down, could you share:

    1. The OpenScreen version and how you installed it (AppImage, pacman, Nix…).
    2. Your display resolution and scaling factor in niri.
    3. The export settings you used (resolution and quality preset).
    4. Whether the blur shows on the whole video or only on zoomed-in parts.
    5. If possible, the ffprobe output of the exported file.

  3. dealenx commented on Nov 2, 2026

    @dealenx
    dealenx
    on Nov 2, 2026
    Author
    1. pacman, download from release build
    2. 1920x1080
    3. default
    4. whole video when screen animation happens

    (Please ignore the microphone sound; I just grabbed the first one I found):

    Recording_02_11_2026_17_16_30.mp4

  4. EtienneLescot commented on Nov 5, 2026

    @EtienneLescot
    EtienneLescot
    on Nov 5, 2026
    Member

    Thanks a lot for the video, it helped! Looking at it frame by frame, this isn't blur from a low bitrate: it's stale blocks of older content left in the picture while the page scrolls, and they're already in the source recording, not added by the export.

    Could you try one quick test? Launch OpenScreen from a terminal with:

    OPENSCREEN_LINUX_ENCODER=software openscreen

    record the same scroll, and tell us whether the ghosting is gone. If you can, also attach the raw recording (the .mp4 OpenScreen saves before export, in its recordings folder) from both runs. And which GPU are you on?

  5. dealenx commented on Nov 6, 2026

    @dealenx
    dealenx
    on Nov 6, 2026
    Author

    I want to try again next week ; I’m really busy this week :(

  6. dealenx commented on Nov 10, 2026

    @dealenx
    dealenx
    on Nov 10, 2026
    Author

    version 2.1.2

    I have found time to test with env arg OPENSCREEN_LINUX_ENCODER=software and I've started using the command:

    OPENSCREEN_LINUX_ENCODER=software openscreen

    How You can see it is helpful for Wayland / Niri on CachyOS!

    Recording_10_11_2026_12_07_17.mp4

    UPD:
    I want to record some videos and I'll check efficiency of openscreen with this env setup

  7. dealenx commented on Nov 10, 2026

    @dealenx
    dealenx
    on Nov 10, 2026
    Author
    Recording_10_11_2026_16_13_51.mp4

    v2.2.2 but it is working normal only with OPENSCREEN_LINUX_ENCODER=software

  8. EtienneLescot commented on Nov 10, 2026

    @EtienneLescot
    EtienneLescot
    on Nov 10, 2026
    Member

    Thanks a lot for testing on both versions, @dealenx, and for the videos! That settles where it comes from: the ghosting only happens on the hardware path, where OpenScreen reads niri's GPU buffers directly on NVIDIA. We're now looking for the fix there.

    In the meantime, OPENSCREEN_LINUX_ENCODER=software openscreen is a safe way to keep recording.

  9. EtienneLescot commented on Nov 10, 2026

    @EtienneLescot
    EtienneLescot
    on Nov 10, 2026
    Member

    Thanks again for the tests, @dealenx! We read the hardware path end to end, and your laptop has two GPUs (the Radeon inside the Ryzen 5 3550H, next to the GTX 1650), which changes the picture. Two things would tell us exactly where the stale blocks come from, with no test build needed:

    1. Start a take that ghosts with plain openscreen from a terminal, and share the [capture-linux] lines it prints (especially dmabuf-import, encoder-selection and buffer-info; the file from tray → Save Diagnostics works too), plus the output of: 
      ls -l /dev/dri/by-path
    2. Record the same scroll with: 
      OPENSCREEN_LINUX_ENCODER=vulkan openscreen
      and tell us whether the ghosting is still there. This keeps a GPU encoder but turns off the direct read of niri's GPU buffers, so it separates the two suspects.

    OPENSCREEN_LINUX_ENCODER=software openscreen stays the safe way to record in the meantime.

  10. dealenx commented on Nov 10, 2026

    @dealenx
    dealenx
    on Nov 10, 2026
    Author
    Recording_11_11_2026_01_49_43.mp4 
    
~
❯ ls -l /dev/dri/by-path
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:01:00.0-card -> ../card1
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:01:00.0-render -> ../renderD128
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:05:00.0-card -> ../card2
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:05:00.0-render -> ../renderD129

~
❯ lspci -nn | grep -i vga
01:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU117M [GeForce GTX 1650 Mobile / Max-Q] [10de:1f91] (rev a1)
05:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Picasso/Raven 2 [Radeon Vega Series / Radeon Vega Mobile Series] [1002:15d8] (rev c2)

~
❯ ls -l /dev/dri/by-path
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:01:00.0-card -> ../card1
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:01:00.0-render -> ../renderD128
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:05:00.0-card -> ../card2
lrwxrwxrwx - root 10 ноя 23:31  pci-0000:05:00.0-render -> ../renderD129

~
❯ lspci -nn | grep -i vga
01:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU117M [GeForce GTX 1650 Mobile / Max-Q] [10de:1f91] (rev a1)
05:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Picasso/Raven 2 [Radeon Vega Series / Radeon Vega Mobile Series] [1002:15d8] (rev c2)

~

    
                             dealenx@cachyos-asus-tuf-x8664
       ...........           ------------------------------
      /,,,,...../              OS  CachyOS x86_64
     /,,,..,,../   ()           Kernel  Linux 7.2.9-1-cachyos
    /,,,,../                   Uptime  2 hours, 17 mins
   /,,,.../     /'\            Packages  1309 (pacman)
  /,...../      \,/            Shell  fish 4.9.3
  \,,,,,,\\             _      WM  niri 26.04 (Wayland)
   \,,....\\           / \      Term  alacritty 0.17.0
    \...,..\\          \_/     󰻠 CPU  AMD Ryzen 5 3550H (8) @ 3.70 GHz
     \..,,,,,,,,,,,,/          󰢮 GPU  NVIDIA GeForce GTX 1650 Mobile / Max-Q [Discrete]
      \,,........,,/           󰍛 Memory  5.07 GiB / 29.30 GiB (17%)
                               󰓡 Swap  1.73 MiB / 29.30 GiB (0%)
                               󰋊 Disk (/)  441.67 GiB / 472.94 GiB (93%) - btrfs



                             ------------------------------
~
❯ OPENSCREEN_LINUX_ENCODER=vulkan openscreen
RECORDINGS_DIR: /home/dealenx/.config/openscreen/recordings
User Data Path: /home/dealenx/.config/openscreen
Global shortcut registered: CommandOrControl+Shift+O
[23874:1111/014850.439332:ERROR:components/dbus/xdg/request.cc:169] Request ended (non-user cancelled).
[23874:1111/014850.439362:ERROR:ui/base/accelerators/global_accelerator_listener/global_accelerator_listener_linux.cc:389] Failed to call BindShortcuts (error code 5).
[capture-linux] [AVHWFramesContext @ 0x55f6bb46a9c0] Unsupported format: bgr0.
[capture-linux] source selected {"sourceKind":"monitor"}
[native-linux] starting capture {
  outputPath: '/home/dealenx/.config/openscreen/recordings/recording-1794336553428.mp4',
  prepared: true,
  cursor: { mode: 'system' },
  audio: {
    system: { enabled: false },
    microphone: {
      enabled: false,
      deviceName: 'USB Condenser Microphone Моно',
      gain: 1
    }
  },
  video: { fps: 60 }
}
[capture-linux] portal stream started {"width":1920,"height":1080,"sourceKind":"monitor"}
[capture-linux] radv: RADV_PERFTEST=video_encode is deprecated and will be removed in future Mesa releases. Please use RADV_EXPERIMENTAL=video_encode instead.
[capture-linux] encoder {"video":"vulkan","rejected":[]}
[capture-linux] capture started {"width":1920,"height":1080,"fps":60}
[capture-linux] [mp4 @ 0x55f6bb28f480] Estimating the duration of the last packet in a fragment, consider setting the duration field in AVPacket instead.
[capture-linux] frames-dropped: 158 captured frame(s) were replaced before the encoder could take them, so the recording holds an older picture across those moments. The machine could not keep up with the capture rate.
[capture-linux] capture stopped {"frames":214,"dropped":158,"durationMs":22533,"convertMs":3.6536404345794393,"uploadMs":0.8411837710280374,"encodeMs":6.678044364485981}
[native-linux] capture stored {
  path: '/home/dealenx/.config/openscreen/recordings/recording-1794336553428.mp4',
  frames: 214,
  droppedFrames: 158,
  durationMs: 22533,
  videoEncoder: 'vulkan',
  cursorSamples: 0
}
[d3d] adaptateur Vulkan : NVIDIA GeForce GTX 1650 (DiscreteGpu, Vulkan) -> backend Hardware, export dmabuf actif
[d3d] adaptateur Vulkan : NVIDIA GeForce GTX 1650 (DiscreteGpu, Vulkan) -> backend Hardware, export dmabuf actif
[whisper-stt-server] [whisper-stt] boot: model=/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-small-q8_0.bin vad_model=/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-silero-v6.2.0.bin host=127.0.0.1 port=33673 threads=8
whisper_init_from_file_with_params_no_state: loading model from '/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-small-q8_0.bin'
whisper_init_with_params_no_state: use gpu    = 1
whisper_init_with_params_no_state: flash attn = 0
whisper_init_with_params_no_state: gpu_device = 0
whisper_init_with_params_no_state: dtw        = 1
[whisper-stt-server] ggml_vulkan: Found 2 Vulkan devices:
[whisper-stt-server] ggml_vulkan: 0 = AMD Radeon Vega 8 Graphics (RADV RAVEN) (radv) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 65536 | int dot: 0 | matrix cores: none
[whisper-stt-server] ggml_vulkan: 1 = NVIDIA GeForce GTX 1650 (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 0 | matrix cores: none
[whisper-stt-server] whisper_init_with_params_no_state: devices    = 3
whisper_init_with_params_no_state: backends   = 2
whisper_model_load: loading model
whisper_model_load: n_vocab       = 51865
whisper_model_load: n_audio_ctx   = 1500
whisper_model_load: n_audio_state = 768
whisper_model_load: n_audio_head  = 12
whisper_model_load: n_audio_layer = 12
whisper_model_load: n_text_ctx    = 448
whisper_model_load: n_text_state  = 768
whisper_model_load: n_text_head   = 12
whisper_model_load: n_text_layer  = 12
whisper_model_load: n_mels        = 80
whisper_model_load: ftype         = 7
whisper_model_load: qntvr         = 2
whisper_model_load: type          = 3 (small)
[whisper-stt-server] whisper_model_load: adding 1608 extra tokens
[whisper-stt-server] whisper_model_load: n_langs       = 99
[whisper-stt-server] whisper_model_load:      Vulkan0 total size =   263.87 MB
[whisper-stt-server] whisper_model_load: model size    =  263.87 MB
[whisper-stt-server] whisper_backend_init_gpu: device 0: Vulkan0 (type: 2)
whisper_backend_init_gpu: found GPU device 0: Vulkan0 (type: 2, cnt: 0)
whisper_backend_init_gpu: using Vulkan0 backend
[whisper-stt-server] whisper_init_state: kv self size  =   18.87 MB
[whisper-stt-server] whisper_init_state: kv cross size =   56.62 MB
[whisper-stt-server] whisper_init_state: kv pad  size  =    4.72 MB
[whisper-stt-server] whisper_init_state: alignment heads masks size = 480 B
[whisper-stt-server] whisper_init_state: compute buffer (conv)   =   23.38 MB
[whisper-stt-server] whisper_init_state: compute buffer (encode) =  132.63 MB
[whisper-stt-server] whisper_init_state: compute buffer (cross)  =   24.63 MB
[whisper-stt-server] whisper_init_state: compute buffer (decode) =  168.84 MB
[whisper-stt] model loaded; backend=whispercpp-vulkan
whisper_vad_init_from_file_with_params: loading VAD model from '/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-silero-v6.2.0.bin'
whisper_vad_init_with_params: model type: silero-16k
whisper_vad_init_with_params: model version: 6.2.0
whisper_vad_init_with_params: n_encoder_layers = 4
whisper_vad_init_with_params: encoder_in_channels[0] = 129
whisper_vad_init_with_params: encoder_in_channels[1] = 128
whisper_vad_init_with_params: encoder_in_channels[2] = 64
whisper_vad_init_with_params: encoder_in_channels[3] = 64
whisper_vad_init_with_params: encoder_out_channels[0] = 128
whisper_vad_init_with_params: encoder_out_channels[1] = 64
whisper_vad_init_with_params: encoder_out_channels[2] = 64
whisper_vad_init_with_params: encoder_out_channels[3] = 128
whisper_vad_init_with_params: lstm_input_size = 128
whisper_vad_init_with_params: lstm_hidden_size = 128
whisper_vad_init_with_params: final_conv_in = 128
whisper_vad_init_with_params: final_conv_out = 1
whisper_vad_init_with_params:          CPU total size =     0.88 MB
whisper_vad_init_with_params: model size    =    0.88 MB
whisper_backend_init_gpu: no GPU found
[whisper-stt-server] whisper_vad_init_context: compute buffer (VAD)   =    1.60 MB
[whisper-stt] VAD model loaded; path=/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-silero-v6.2.0.bin
[whisper-stt-server] [whisper-stt] listening on 127.0.0.1:33673
Error occurred in handler for 'stt:transcribe': qut [NoAudioTrackError]: No decodable audio in /home/dealenx/.config/openscreen/recordings/recording-1794336553428.mp4: [out#0/f32le @ 0x558e5c88f380] Output file does not contain any stream
Error opening output file -.
Error opening output files: Invalid argument
    at ChildProcess.<anonymous> (file:///opt/Openscreen/resources/app.asar/dist-electron/main-BOoZm-yU.js:116516:55)
    at Object.onceWrapper (node:events:623:26)
    at ChildProcess.emit (node:events:508:28)
    at maybeClose (node:internal/child_process:1100:16)
    at Socket.<anonymous> (node:internal/child_process:457:11)
    at Socket.emit (node:events:508:28)
    at Pipe.<anonymous> (node:net:346:12)
[d3d] adaptateur Vulkan : NVIDIA GeForce GTX 1650 (DiscreteGpu, Vulkan) -> backend Hardware, export dmabuf actif
[h264_vaapi @ 0x17ec0fc16c00] No usable encoding entrypoint found for profile VAProfileH264High (7).
[libopenh264 @ 0x17ec0fc16800] [OpenH264] this = 0x0x17ec042e3240, Warning:bEnableFrameSkip = 0,bitrate can't be controlled for RC_QUALITY_MODE,RC_BITRATE_MODE and RC_TIMESTAMP_MODE without enabling skip frame.
[pipeline] encodeur video : libopenh264 (software YUV420P)
[pipeline] encodeur video : software
[pipeline] warning: clip #0 déclaré audio mais sans flux décodable; silence conservé
[aac @ 0x17ec0fc17400] Qavg: 65536.000

  11. dealenx commented on Nov 10, 2026

    @dealenx
    dealenx
    on Nov 10, 2026
    Author
    Recording_11_11_2026_02_00_06.mp4 
                                dealenx@cachyos-asus-tuf-x8664
       ...........           ------------------------------
      /,,,,...../              OS  CachyOS x86_64
     /,,,..,,../   ()           Kernel  Linux 7.2.9-1-cachyos
    /,,,,../                   Uptime  2 hours, 28 mins
   /,,,.../     /'\            Packages  1309 (pacman)
  /,...../      \,/            Shell  fish 4.9.3
  \,,,,,,\\             _      WM  niri 26.04 (Wayland)
   \,,....\\           / \      Term  alacritty 0.17.0
    \...,..\\          \_/     󰻠 CPU  AMD Ryzen 5 3550H (8) @ 3.70 GHz
     \..,,,,,,,,,,,,/          󰢮 GPU  NVIDIA GeForce GTX 1650 Mobile / Max-Q [Discrete]
      \,,........,,/           󰍛 Memory  5.63 GiB / 29.30 GiB (19%)
                               󰓡 Swap  1.71 MiB / 29.30 GiB (0%)
                               󰋊 Disk (/)  441.73 GiB / 472.94 GiB (93%) - btrfs



                             ------------------------------
~
❯ env -u OPENSCREEN_LINUX_ENCODER openscreen
RECORDINGS_DIR: /home/dealenx/.config/openscreen/recordings
User Data Path: /home/dealenx/.config/openscreen
Global shortcut registered: CommandOrControl+Shift+O
[25926:1111/015917.963833:ERROR:components/dbus/xdg/request.cc:169] Request ended (non-user cancelled).
[25926:1111/015917.963882:ERROR:ui/base/accelerators/global_accelerator_listener/global_accelerator_listener_linux.cc:389] Failed to call BindShortcuts (error code 5).
[app] quitting

~
❯ env -u OPENSCREEN_LINUX_ENCODER openscreen
RECORDINGS_DIR: /home/dealenx/.config/openscreen/recordings
User Data Path: /home/dealenx/.config/openscreen
Global shortcut registered: CommandOrControl+Shift+O
[26110:1111/015936.969322:ERROR:components/dbus/xdg/request.cc:169] Request ended (non-user cancelled).
[26110:1111/015936.969359:ERROR:ui/base/accelerators/global_accelerator_listener/global_accelerator_listener_linux.cc:389] Failed to call BindShortcuts (error code 5).
[capture-linux] [AVHWFramesContext @ 0x56506266ab00] Unsupported format: bgr0.
[capture-linux] source selected {"sourceKind":"monitor"}
[native-linux] starting capture {
  outputPath: '/home/dealenx/.config/openscreen/recordings/recording-1794337190971.mp4',
  prepared: true,
  cursor: { mode: 'system' },
  audio: {
    system: { enabled: false },
    microphone: {
      enabled: false,
      deviceName: 'USB Condenser Microphone Моно',
      gain: 1
    }
  },
  video: { fps: 60 }
}
[capture-linux] portal stream started {"width":1920,"height":1080,"sourceKind":"monitor"}
[capture-linux] [h264_vaapi @ 0x5650625e7200] No usable encoding entrypoint found for profile VAProfileH264High (7).
[capture-linux] radv: RADV_PERFTEST=video_encode is deprecated and will be removed in future Mesa releases. Please use RADV_EXPERIMENTAL=video_encode instead.
[capture-linux] encoder {"video":"vulkan","rejected":["vaapi: avcodec_open2: Function not implemented"]}
[capture-linux] capture started {"width":1920,"height":1080,"fps":60}
[capture-linux] [mp4 @ 0x5650624b1500] Estimating the duration of the last packet in a fragment, consider setting the duration field in AVPacket instead.
[capture-linux] frames-dropped: 54 captured frame(s) were replaced before the encoder could take them, so the recording holds an older picture across those moments. The machine could not keep up with the capture rate.
[capture-linux] capture stopped {"frames":106,"dropped":54,"durationMs":7916,"convertMs":2.870781283018868,"uploadMs":1.2199449056603773,"encodeMs":4.55837}
[native-linux] capture stored {
  path: '/home/dealenx/.config/openscreen/recordings/recording-1794337190971.mp4',
  frames: 106,
  droppedFrames: 54,
  durationMs: 7916,
  videoEncoder: 'vulkan',
  cursorSamples: 0
}
[d3d] adaptateur Vulkan : NVIDIA GeForce GTX 1650 (DiscreteGpu, Vulkan) -> backend Hardware, export dmabuf actif
[d3d] adaptateur Vulkan : NVIDIA GeForce GTX 1650 (DiscreteGpu, Vulkan) -> backend Hardware, export dmabuf actif
[whisper-stt-server] [whisper-stt] boot: model=/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-small-q8_0.bin vad_model=/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-silero-v6.2.0.bin host=127.0.0.1 port=45097 threads=8
whisper_init_from_file_with_params_no_state: loading model from '/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-small-q8_0.bin'
whisper_init_with_params_no_state: use gpu    = 1
whisper_init_with_params_no_state: flash attn = 0
whisper_init_with_params_no_state: gpu_device = 0
whisper_init_with_params_no_state: dtw        = 1
[whisper-stt-server] ggml_vulkan: Found 2 Vulkan devices:
[whisper-stt-server] ggml_vulkan: 0 = AMD Radeon Vega 8 Graphics (RADV RAVEN) (radv) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 64 | shared memory: 65536 | int dot: 0 | matrix cores: none
[whisper-stt-server] ggml_vulkan: 1 = NVIDIA GeForce GTX 1650 (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 0 | matrix cores: none
[whisper-stt-server] whisper_init_with_params_no_state: devices    = 3
whisper_init_with_params_no_state: backends   = 2
whisper_model_load: loading model
whisper_model_load: n_vocab       = 51865
whisper_model_load: n_audio_ctx   = 1500
whisper_model_load: n_audio_state = 768
whisper_model_load: n_audio_head  = 12
whisper_model_load: n_audio_layer = 12
whisper_model_load: n_text_ctx    = 448
whisper_model_load: n_text_state  = 768
whisper_model_load: n_text_head   = 12
whisper_model_load: n_text_layer  = 12
whisper_model_load: n_mels        = 80
whisper_model_load: ftype         = 7
whisper_model_load: qntvr         = 2
whisper_model_load: type          = 3 (small)
[whisper-stt-server] whisper_model_load: adding 1608 extra tokens
[whisper-stt-server] whisper_model_load: n_langs       = 99
[whisper-stt-server] whisper_model_load:      Vulkan0 total size =   263.87 MB
[whisper-stt-server] whisper_model_load: model size    =  263.87 MB
[whisper-stt-server] whisper_backend_init_gpu: device 0: Vulkan0 (type: 2)
whisper_backend_init_gpu: found GPU device 0: Vulkan0 (type: 2, cnt: 0)
whisper_backend_init_gpu: using Vulkan0 backend
[whisper-stt-server] whisper_init_state: kv self size  =   18.87 MB
[whisper-stt-server] whisper_init_state: kv cross size =   56.62 MB
[whisper-stt-server] whisper_init_state: kv pad  size  =    4.72 MB
[whisper-stt-server] whisper_init_state: alignment heads masks size = 480 B
[whisper-stt-server] whisper_init_state: compute buffer (conv)   =   23.38 MB
[whisper-stt-server] whisper_init_state: compute buffer (encode) =  132.63 MB
[whisper-stt-server] whisper_init_state: compute buffer (cross)  =   24.63 MB
[whisper-stt-server] whisper_init_state: compute buffer (decode) =  168.84 MB
[whisper-stt] model loaded; backend=whispercpp-vulkan
whisper_vad_init_from_file_with_params: loading VAD model from '/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-silero-v6.2.0.bin'
whisper_vad_init_with_params: model type: silero-16k
[whisper-stt-server] whisper_vad_init_with_params: model version: 6.2.0
whisper_vad_init_with_params: n_encoder_layers = 4
whisper_vad_init_with_params: encoder_in_channels[0] = 129
whisper_vad_init_with_params: encoder_in_channels[1] = 128
whisper_vad_init_with_params: encoder_in_channels[2] = 64
whisper_vad_init_with_params: encoder_in_channels[3] = 64
whisper_vad_init_with_params: encoder_out_channels[0] = 128
whisper_vad_init_with_params: encoder_out_channels[1] = 64
whisper_vad_init_with_params: encoder_out_channels[2] = 64
whisper_vad_init_with_params: encoder_out_channels[3] = 128
whisper_vad_init_with_params: lstm_input_size = 128
whisper_vad_init_with_params: lstm_hidden_size = 128
whisper_vad_init_with_params: final_conv_in = 128
whisper_vad_init_with_params: final_conv_out = 1
whisper_vad_init_with_params:          CPU total size =     0.88 MB
[whisper-stt-server] whisper_vad_init_with_params: model size    =    0.88 MB
whisper_backend_init_gpu: no GPU found
[whisper-stt-server] whisper_vad_init_context: compute buffer (VAD)   =    1.60 MB
[whisper-stt] VAD model loaded; path=/home/dealenx/.config/openscreen/stt-models/whisper-ggml/ggml-silero-v6.2.0.bin
[whisper-stt-server] [whisper-stt] listening on 127.0.0.1:45097
Error occurred in handler for 'stt:transcribe': qut [NoAudioTrackError]: No decodable audio in /home/dealenx/.config/openscreen/recordings/recording-1794337190971.mp4: [out#0/f32le @ 0x558bd3655900] Output file does not contain any stream
Error opening output file -.
Error opening output files: Invalid argument
    at ChildProcess.<anonymous> (file:///opt/Openscreen/resources/app.asar/dist-electron/main-BOoZm-yU.js:116516:55)
    at Object.onceWrapper (node:events:623:26)
    at ChildProcess.emit (node:events:508:28)
    at maybeClose (node:internal/child_process:1100:16)
    at Socket.<anonymous> (node:internal/child_process:457:11)
    at Socket.emit (node:events:508:28)
    at Pipe.<anonymous> (node:net:346:12)
[d3d] adaptateur Vulkan : NVIDIA GeForce GTX 1650 (DiscreteGpu, Vulkan) -> backend Hardware, export dmabuf actif
[h264_vaapi @ 0xf640d879c00] No usable encoding entrypoint found for profile VAProfileH264High (7).
[libopenh264 @ 0xf640d52c400] [OpenH264] this = 0x0xf640d575ec0, Warning:bEnableFrameSkip = 0,bitrate can't be controlled for RC_QUALITY_MODE,RC_BITRATE_MODE and RC_TIMESTAMP_MODE without enabling skip frame.
[pipeline] encodeur video : libopenh264 (software YUV420P)
[pipeline] encodeur video : software
[pipeline] warning: clip #0 déclaré audio mais sans flux décodable; silence conservé
[aac @ 0xf640d87a400] Qavg: 65536.000
  12. EtienneLescot
    added a commit that references this issue on Nov 13, 2026

    fix(linux): skip h264_vulkan on RADV Raven, whose takes come out corrupt

    af44d2c

  13. EtienneLescot commented on Nov 13, 2026

    @EtienneLescot
    EtienneLescot
    on Nov 13, 2026
    Member

    @dealenx the fix is merged in #1542: on your Vega 8 iGPU the Linux recorder now skips the Vulkan encoder and records with the software encoder. Could you confirm your takes come out clean on the next release?

  14. dealenx commented on Nov 14, 2026

    @dealenx
    dealenx
    on Nov 14, 2026
    Author

    @EtienneLescot I'll test next release

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions