aboutsummaryrefslogtreecommitdiffstats
path: root/whisper.cpp/README
diff options
context:
space:
mode:
Diffstat (limited to 'whisper.cpp/README')
-rw-r--r--whisper.cpp/README6
1 files changed, 5 insertions, 1 deletions
diff --git a/whisper.cpp/README b/whisper.cpp/README
index 3894eba..74b89f5 100644
--- a/whisper.cpp/README
+++ b/whisper.cpp/README
@@ -5,7 +5,7 @@ recognition (ASR) model, plus NVIDIA Parakeet models.
- AVX intrinsics support for x86 architectures
- Mixed F16 / F32 precision
- Integer quantization support
- - Vulkan GPU backend
+ - Vulkan and OpenVINO GPU backends
- OpenVINO encoder support
Installs whisper-cli, whisper-server, whisper-bench, whisper-quantize,
@@ -19,4 +19,8 @@ is accepted, not only WAV, MP3 and FLAC.
OpenVINO support is built in. The encoder must be converted to OpenVINO
IR format first, see "OpenVINO Support" in the bundled README.md.
+Both GPU backends are built in; whisper uses the first GPU device it
+finds. Pick one with whisper-cli -dev N (the startup log lists the
+devices) and compare the timings it prints at exit.
+
OpenBLAS is autodetected: if installed, it is used for the BLAS backend.