whisper.cpp vs faster-whisper on a Mac
On an Apple Silicon Mac, whisper.cpp is the natural first choice because it runs on the GPU through Metal. faster-whisper is built on CTranslate2, whose hardware page lists CPUs plus NVIDIA and AMD GPUs but no Metal. So its published GPU numbers, measured on an NVIDIA card, don't carry over to a Mac. We use whisper.cpp in FnFlow and haven't run faster-whisper on the same recordings.
At a glance
| whisper.cpp | faster-whisper | |
|---|---|---|
| Language | C/C++ | Python on CTranslate2 |
| Apple GPU | Yes, runs fully on the GPU through Metal | Not listed |
| NVIDIA GPU | Supported | Yes, needs CUDA 12 libraries |
| Neural Engine | Optional Core ML encoder, over 3x faster than CPU-only per its README | – |
| Quantization | q5, q8 and others | 8-bit on CPU and GPU |
| Install | Homebrew or CMake | pip |
| Keeps a model warm | whisper-server, if you build from source | In your own process |
From github.com/ggml-org/whisper.cpp, github.com/SYSTRAN/faster-whisper and opennmt.net/CTranslate2, checked 6 November 2026. Compiling CTranslate2 yourself can change its hardware support.
What we measured on whisper.cpp
On an M4 Pro with Metal, large-v3-turbo q8_0 took a median of 1.98 s per dictation on 5 to 18 second recordings, including loading the model from disk every time. With the model kept loaded in FnFlow, a 7.6 s sentence comes back in about 1.2 s. Full numbers and commands are in our benchmark.
Measure it yourself
Same file, same model size, and throw away the first run of each, since it includes downloads and Metal compilation:
time whisper-cli -m ggml-large-v3-turbo-q8_0.bin -f test.wav -l auto
time python3 -c "from faster_whisper import WhisperModel as W; \
s,_=W('large-v3-turbo',device='cpu',compute_type='int8').transcribe('test.wav'); \
print(' '.join(x.text for x in s))"
faster-whisper returns a generator, so the transcription only happens when you read the segments. The print at the end is what makes the timing meaningful.
Recommendation
For anything that lives on a Mac, start with whisper.cpp. Setup is in how to run Whisper locally on a Mac. Stick with faster-whisper if you're already on it in Python and the CPU speed is good enough. And if you only want to dictate, you don't need either: FnFlow bundles whisper.cpp and keeps the model warm for you.
Questions
Does faster-whisper use the Mac GPU?
CTranslate2's hardware support page lists CPUs, NVIDIA GPUs and AMD ROCm, with nothing for Apple's GPU. That page covers prebuilt binaries. Expect CPU inference on a Mac.
Is faster-whisper faster than whisper.cpp?
Its README says up to 4 times faster than OpenAI's Python Whisper, with GPU benchmarks on an RTX 3070 Ti. That's a comparison on different software and hardware, so it doesn't answer this question for a Mac. Time both on your machine.
Which is easier to use from Python?
faster-whisper. It's a pip package with a Python API. whisper.cpp is a C/C++ project with a CLI and server, though third-party bindings exist.
When would I choose faster-whisper on a Mac?
When your code is in Python and also has to run on NVIDIA servers, so that you keep a single codebase, or when you depend on its built-in voice activity detection and batching.
Local Wispr Flow alternativeFnFlow vs VoiceInkVoice input for CursorHow 7 Mac dictation apps get you to your first dictationWhisper.cpp on Apple Silicon: which large-v3-turbo quantization to shipHow I built a local dictation appVoice dictation in Claude Code and the terminalWhy holding fn doesn't work for dictation on Mac (and how to fix it)How to run Whisper locally on an Apple Silicon MacSpokenly alternative for developersFnFlow vs EnviousWisprHow to dictate a good prompt for AI coding toolsVoice dictation in VS Code on a MacTeaching Whisper your jargon: initial prompt and replacementsWhy Whisper types "Thank you" on silence, and how to filter itPrivate dictation on a Mac at work: what actually leaves your machineVoice input for Codex CLI and the ChatGPT appVoice coding on a Mac: what actually worksWriting git commit messages by voiceDictating code review comments on GitHubDictation in IntelliJ IDEA, PyCharm and WebStorm on a MacDictation in Zed on a MacDictation in Terminal, iTerm2 and WarpWhich Whisper model to use on an Apple Silicon MacPush-to-talk dictation on a Mac: hold, toggle or streamSuperwhisper alternative for developersClaude Code voice mode: how to enable /voice and fix itWhisper apps: five ways to run Whisper locallyTalon voice coding vs a dictation appPush to talk on a MacDoes Claude have a voice mode?VoiceInk vs Wispr FlowVoiceInk vs SuperwhisperParakeet vs Whisper on a MacApple SpeechAnalyzer vs WhisperMlx-whisper vs whisper.cppGitHub Copilot voice inputGemini CLI voice inputVoice input in Google AntigravityVibe coding with voice input