From 3ef918b667f71ad838e430ae540f32c1b42e2d2c Mon Sep 17 00:00:00 2001 From: "Danilo M." Date: Fri, 25 Sep 2026 19:35:12 +0200 Subject: whisper.cpp: enable Vulkan, SDL2 and ffmpeg Groundwork for a local voice assistant. SDL2 adds the microphone tools: whisper-stream for live transcription, whisper-command for voice commands, plus whisper-talk-llama and whisper-lsp. ffmpeg lets the tools decode any audio format ffmpeg reads, not only the WAV, MP3 and FLAC that the bundled miniaudio handles. Vulkan runs the whole model on the GPU; OpenVINO still covers the encoder when a converted encoder model is present. All three come from the -current base, so REQUIRES is unchanged. Test-built on -current, sbopkglint clean. Co-Authored-By: Claude Opus 5.5 --- whisper.cpp/whisper.cpp.SlackBuild | 3 +++ 1 file changed, 3 insertions(+) (limited to 'whisper.cpp/whisper.cpp.SlackBuild') diff --git a/whisper.cpp/whisper.cpp.SlackBuild b/whisper.cpp/whisper.cpp.SlackBuild index 76893eb..f4de5fe 100644 --- a/whisper.cpp/whisper.cpp.SlackBuild +++ b/whisper.cpp/whisper.cpp.SlackBuild @@ -100,6 +100,9 @@ cd build -DWHISPER_ALL_WARNINGS=OFF \ -DGGML_BUILD_EXAMPLES=OFF \ -DGGML_BUILD_TESTS=OFF \ + -DGGML_VULKAN=ON \ + -DWHISPER_SDL2=ON \ + -DWHISPER_COMMON_FFMPEG=ON \ $BLAS \ -DWHISPER_OPENVINO=ON \ -DOpenVINO_DIR=$OVDIR/cmake \ -- cgit v1.2.3