whisper.cpp

High-performance inference of OpenAI's Whisper automatic speech recognition (ASR) model:

• Plain C/C++ implementation without dependencies
• Apple silicon first-class citizen - optimized via Arm Neon and Accelerate framework
• AVX intrinsics support for x86 architectures
• Mixed F16 / F32 precision
• Low memory usage (Flash Attention + Flash Forward)
• Zero memory allocations at runtime
• Runs on the CPU
• C-style API

Supported platforms:

• Mac OS (Intel and Arm)
• iOS
• Linux
• WebAssembly
• Windows (MSVC and MinGW)
• Raspberry Pi
• Android

The entire implementation of the model is contained in 2 source files:

• Tensor operations: ggml.h / ggml.c
• Transformer inference: whisper.h / whisper.cpp

Having such a lightweight implementation of the model allows to easily integrate it in different platforms and applications. As an example, here is a video of running the model on an iPhone 13 device - fully offline, on-device:

https://user-images.githubusercontent.com/1991296/197385372-962a6dea-bca1-4d50-bf96-1d8c27b98c81.mp4

Implementation details

• The core tensor operations are implemented in C (ggml.h / ggml.c)
• The transformer model and the high-level C-style API are implemented in C++ (whisper.h / whisper.cpp)
• Sample usage is demonstrated in main.cpp
• Sample real-time audio transcription from the microphone is demonstrated in stream.cpp
• Various other examples are available in the examples folder

The tensor operators are optimized heavily for Apple silicon CPUs. Depending on the computation size, Arm Neon SIMD instrisics or CBLAS Accelerate framework routines are used. The latter are especially effective for bigger sizes since the Accelerate framework utilizes the special-purpose AMX coprocessor available in modern Apple products.

Limitations

• Inference only

• No GPU support

• Very basic greedy sampling scheme - always pick up the token with highest probability. This should be similar to the GreedyDecoder from the original python implementation, so in order to make a fair comparison between the 2 implementations, make sure to run the python code with the following parameters:

  whisper --best_of None --beam_size None ...
  

  In the future, whisper.cpp will support more sampling strategies.

Quick start

First, download one of the Whisper models converted in ggml format. For example:

bash ./models/download-ggml-model.sh base.en

Now build the main example and transcribe an audio file like this:

# build the main example
make

# transcribe an audio file
./main -f input.wav

---

For a quick demo, simply run make base.en:

$ make base.en

cc  -I.              -O3 -std=c11   -pthread -DGGML_USE_ACCELERATE   -c ggml.c -o ggml.o
c++ -I. -I./examples -O3 -std=c++11 -pthread -c whisper.cpp -o whisper.o
c++ -I. -I./examples -O3 -std=c++11 -pthread examples/main/main.cpp whisper.o ggml.o -o main  -framework Accelerate
./main -h

usage: ./main [options] file0.wav file1.wav ...

options:
  -h,       --help           show this help message and exit
  -s SEED,  --seed SEED      RNG seed (default: -1)
  -t N,     --threads N      number of threads to use during computation (default: 4)
  -p N,     --processors N   number of processors to use during computation (default: 1)
  -ot N,    --offset-t N     time offset in milliseconds (default: 0)
  -on N,    --offset-n N     segment index offset (default: 0)
  -mc N,    --max-context N  maximum number of text context tokens to store (default: max)
  -ml N,    --max-len N      maximum segment length in characters (default: 0)
  -wt N,    --word-thold N   word timestamp probability threshold (default: 0.010000)
  -v,       --verbose        verbose output
            --translate      translate from source language to english
  -otxt,    --output-txt     output result in a text file
  -ovtt,    --output-vtt     output result in a vtt file
  -osrt,    --output-srt     output result in a srt file
  -owts,    --output-words   output script for generating karaoke video
  -ps,      --print_special  print special tokens
  -pc,      --print_colors   print colors
  -nt,      --no_timestamps  do not print timestamps
  -l LANG,  --language LANG  spoken language (default: en)
  -m FNAME, --model FNAME    model path (default: models/ggml-base.en.bin)
  -f FNAME, --file FNAME     input WAV file path

bash ./models/download-ggml-model.sh base.en
Downloading ggml model base.en ...
ggml-base.en.bin               100%[========================>] 141.11M  6.34MB/s    in 24s
Done! Model 'base.en' saved in 'models/ggml-base.en.bin'
You can now use it like this:

  $ ./main -m models/ggml-base.en.bin -f samples/jfk.wav


===============================================
Running base.en on all samples in ./samples ...
===============================================

----------------------------------------------
[+] Running base.en on samples/jfk.wav ... (run 'ffplay samples/jfk.wav' to listen)
----------------------------------------------

whisper_model_load: loading model from 'models/ggml-base.en.bin'
whisper_model_load: n_vocab       = 51864
whisper_model_load: n_audio_ctx   = 1500
whisper_model_load: n_audio_state = 512
whisper_model_load: n_audio_head  = 8
whisper_model_load: n_audio_layer = 6
whisper_model_load: n_text_ctx    = 448
whisper_model_load: n_text_state  = 512
whisper_model_load: n_text_head   = 8
whisper_model_load: n_text_layer  = 6
whisper_model_load: n_mels        = 80
whisper_model_load: f16           = 1
whisper_model_load: type          = 2
whisper_model_load: mem_required  = 670.00 MB
whisper_model_load: adding 1607 extra tokens
whisper_model_load: ggml ctx size = 140.60 MB
whisper_model_load: memory size =    22.83 MB
whisper_model_load: model size  =   140.54 MB

system_info: n_threads = 4 / 10 | AVX2 = 0 | AVX512 = 0 | NEON = 1 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 |

main: processing 'samples/jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 processors, lang = en, task = transcribe, timestamps = 1 ...


[00:00:00.000 --> 00:00:11.000]   And so my fellow Americans, ask not what your country can do for you, ask what you can do for your country.


whisper_print_timings:     load time =   105.91 ms
whisper_print_timings:      mel time =    24.62 ms
whisper_print_timings:   sample time =     3.63 ms
whisper_print_timings:   encode time =   324.71 ms / 54.12 ms per layer
whisper_print_timings:   decode time =    83.58 ms / 13.93 ms per layer
whisper_print_timings:    total time =   542.81 ms

The command downloads the base.en model converted to custom ggml format and runs the inference on all .wav samples in the folder samples.

For detailed usage instructions, run: ./main -h

Note that the main example currently runs only with 16-bit WAV files, so make sure to convert your input before running the tool. For example, you can use ffmpeg like this:

ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le output.wav

More audio samples

If you want some extra audio samples to play with, simply run:

make samples

This will download a few more audio files from Wikipedia and convert them to 16-bit WAV format via ffmpeg.

You can download and run the other models as follows:

make tiny.en
make tiny
make base.en
make base
make small.en
make small
make medium.en
make medium
make large

Memory usage

Model	Disk	Mem	SHA
tiny	75 MB	~390 MB	bd577a113a864445d4c299885e0cb97d4ba92b5f
base	142 MB	~500 MB	465707469ff3a37a2b9b8d8f89f2f99de7299dac
small	466 MB	~1.0 GB	55356645c2b361a969dfd0ef2c5a50d530afd8d5
medium	1.5 GB	~2.6 GB	fd9727b6e1217c2f614f9b698455c4ffd82463b4
large	2.9 GB	~4.7 GB	b1caaf735c4cc1429223d5a74f0f4d0b9b59a299

Another example

Here is another example of transcribing a 3:24 min speech in about half a minute on a MacBook M1 Pro, using medium.en model:

Expand to see the result

$ ./main -m models/ggml-medium.en.bin -f samples/gb1.wav -t 8

whisper_model_load: loading model from 'models/ggml-medium.en.bin'
whisper_model_load: n_vocab       = 51864
whisper_model_load: n_audio_ctx   = 1500
whisper_model_load: n_audio_state = 1024
whisper_model_load: n_audio_head  = 16
whisper_model_load: n_audio_layer = 24
whisper_model_load: n_text_ctx    = 448
whisper_model_load: n_text_state  = 1024
whisper_model_load: n_text_head   = 16
whisper_model_load: n_text_layer  = 24
whisper_model_load: n_mels        = 80
whisper_model_load: f16           = 1
whisper_model_load: type          = 4
whisper_model_load: mem_required  = 2610.00 MB
whisper_model_load: adding 1607 extra tokens
whisper_model_load: ggml ctx size = 1644.97 MB
whisper_model_load: memory size =   182.62 MB
whisper_model_load: model size  =  1462.12 MB

main: processing 'samples/gb1.wav' (3179750 samples, 198.7 sec), 8 threads, lang = en, task = transcribe, timestamps = 1 ...

[00:00.000 --> 00:08.000]   My fellow Americans, this day has brought terrible news and great sadness to our country.
[00:08.000 --> 00:17.000]   At nine o'clock this morning, Mission Control in Houston lost contact with our Space Shuttle Columbia.
[00:17.000 --> 00:23.000]   A short time later, debris was seen falling from the skies above Texas.
[00:23.000 --> 00:29.000]   The Columbia's lost. There are no survivors.
[00:29.000 --> 00:32.000]   On board was a crew of seven.
[00:32.000 --> 00:39.000]   Colonel Rick Husband, Lieutenant Colonel Michael Anderson, Commander Laurel Clark,
[00:39.000 --> 00:48.000]   Captain David Brown, Commander William McCool, Dr. Kultna Shavla, and Ilan Ramon,
[00:48.000 --> 00:52.000]   a colonel in the Israeli Air Force.
[00:52.000 --> 00:58.000]   These men and women assumed great risk in the service to all humanity.
[00:58.000 --> 01:03.000]   In an age when space flight has come to seem almost routine,
[01:03.000 --> 01:07.000]   it is easy to overlook the dangers of travel by rocket
[01:07.000 --> 01:12.000]   and the difficulties of navigating the fierce outer atmosphere of the Earth.
[01:12.000 --> 01:18.000]   These astronauts knew the dangers, and they faced them willingly,
[01:18.000 --> 01:23.000]   knowing they had a high and noble purpose in life.
[01:23.000 --> 01:31.000]   Because of their courage and daring and idealism, we will miss them all the more.
[01:31.000 --> 01:36.000]   All Americans today are thinking as well of the families of these men and women
[01:36.000 --> 01:40.000]   who have been given this sudden shock and grief.
[01:40.000 --> 01:45.000]   You're not alone. Our entire nation grieves with you,
[01:45.000 --> 01:52.000]   and those you love will always have the respect and gratitude of this country.
[01:52.000 --> 01:56.000]   The cause in which they died will continue.
[01:56.000 --> 02:04.000]   Mankind is led into the darkness beyond our world by the inspiration of discovery
[02:04.000 --> 02:11.000]   and the longing to understand. Our journey into space will go on.
[02:11.000 --> 02:16.000]   In the skies today, we saw destruction and tragedy.
[02:16.000 --> 02:22.000]   Yet farther than we can see, there is comfort and hope.
[02:22.000 --> 02:29.000]   In the words of the prophet Isaiah, "Lift your eyes and look to the heavens
[02:29.000 --> 02:35.000]   who created all these. He who brings out the starry hosts one by one
[02:35.000 --> 02:39.000]   and calls them each by name."
[02:39.000 --> 02:46.000]   Because of His great power and mighty strength, not one of them is missing.
[02:46.000 --> 02:55.000]   The same Creator who names the stars also knows the names of the seven souls we mourn today.
[02:55.000 --> 03:01.000]   The crew of the shuttle Columbia did not return safely to earth,
[03:01.000 --> 03:05.000]   yet we can pray that all are safely home.
[03:05.000 --> 03:13.000]   May God bless the grieving families, and may God continue to bless America.
[03:13.000 --> 03:41.000]   Audio


whisper_print_timings:     load time =   575.92 ms
whisper_print_timings:      mel time =   230.60 ms
whisper_print_timings:   sample time =    73.19 ms
whisper_print_timings:   encode time = 19552.61 ms / 814.69 ms per layer
whisper_print_timings:   decode time = 13249.96 ms / 552.08 ms per layer
whisper_print_timings:    total time = 33686.27 ms

Real-time audio input example

This is a naive example of performing real-time inference on audio from your microphone. The stream tool samples the audio every half a second and runs the transcription continously. More info is available in issue #10.

./stream -m ./models/ggml-base.en.bin -t 8 --step 500 --length 5000

https://user-images.githubusercontent.com/1991296/194935793-76afede7-cfa8-48d8-a80f-28ba83be7d09.mp4

Confidence color-coding

Adding the --print-colors argument will print the transcribed text using an experimental color coding strategy to highlight words with high or low confidence:

Controlling the length of the generated text segments (experimental)

For example, to limit the line length to a maximum of 16 characters, simply add -ml 16:

./main -m ./models/ggml-base.en.bin -f ./samples/jfk.wav -ml 16

whisper_model_load: loading model from './models/ggml-base.en.bin'
...
system_info: n_threads = 4 / 10 | AVX2 = 0 | AVX512 = 0 | NEON = 1 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 | 

main: processing './samples/jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 processors, lang = en, task = transcribe, timestamps = 1 ...

[00:00:00.000 --> 00:00:00.850]   And so my
[00:00:00.850 --> 00:00:01.590]   fellow
[00:00:01.590 --> 00:00:04.140]   Americans, ask
[00:00:04.140 --> 00:00:05.660]   not what your
[00:00:05.660 --> 00:00:06.840]   country can do
[00:00:06.840 --> 00:00:08.430]   for you, ask
[00:00:08.430 --> 00:00:09.440]   what you can do
[00:00:09.440 --> 00:00:10.020]   for your
[00:00:10.020 --> 00:00:11.000]   country.

Word-level timestamp

The --max-len argument can be used to obtain word-level timestamps. Simply use -ml 1:

./main -m ./models/ggml-base.en.bin -f ./samples/jfk.wav -ml 1

whisper_model_load: loading model from './models/ggml-base.en.bin'
...
system_info: n_threads = 4 / 10 | AVX2 = 0 | AVX512 = 0 | NEON = 1 | FP16_VA = 1 | WASM_SIMD = 0 | BLAS = 1 | 

main: processing './samples/jfk.wav' (176000 samples, 11.0 sec), 4 threads, 1 processors, lang = en, task = transcribe, timestamps = 1 ...

[00:00:00.000 --> 00:00:00.320]  
[00:00:00.320 --> 00:00:00.370]   And
[00:00:00.370 --> 00:00:00.690]   so
[00:00:00.690 --> 00:00:00.850]   my
[00:00:00.850 --> 00:00:01.590]   fellow
[00:00:01.590 --> 00:00:02.850]   Americans
[00:00:02.850 --> 00:00:03.300]  ,
[00:00:03.300 --> 00:00:04.140]   ask
[00:00:04.140 --> 00:00:04.990]   not
[00:00:04.990 --> 00:00:05.410]   what
[00:00:05.410 --> 00:00:05.660]   your
[00:00:05.660 --> 00:00:06.260]   country
[00:00:06.260 --> 00:00:06.600]   can
[00:00:06.600 --> 00:00:06.840]   do
[00:00:06.840 --> 00:00:07.010]   for
[00:00:07.010 --> 00:00:08.170]   you
[00:00:08.170 --> 00:00:08.190]  ,
[00:00:08.190 --> 00:00:08.430]   ask
[00:00:08.430 --> 00:00:08.910]   what
[00:00:08.910 --> 00:00:09.040]   you
[00:00:09.040 --> 00:00:09.320]   can
[00:00:09.320 --> 00:00:09.440]   do
[00:00:09.440 --> 00:00:09.760]   for
[00:00:09.760 --> 00:00:10.020]   your
[00:00:10.020 --> 00:00:10.510]   country
[00:00:10.510 --> 00:00:11.000]  .

Karaoke-style movie generation (experimental)

The main example provides support for output of karaoke-style movies, where the currently pronounced word is highlighted. Use the -wts argument and run the generated bash script. This requires to have ffmpeg installed.

Here are a few "typical" examples:

./main -m ./models/ggml-base.en.bin -f ./samples/jfk.wav -owts
source ./samples/jfk.wav.wts
ffplay ./samples/jfk.wav.mp4

https://user-images.githubusercontent.com/1991296/199337465-dbee4b5e-9aeb-48a3-b1c6-323ac4db5b2c.mp4

---

./main -m ./models/ggml-base.en.bin -f ./samples/mm0.wav -owts
source ./samples/mm0.wav.wts
ffplay ./samples/mm0.wav.mp4

https://user-images.githubusercontent.com/1991296/199337504-cc8fd233-0cb7-4920-95f9-4227de3570aa.mp4

---

./main -m ./models/ggml-base.en.bin -f ./samples/gb0.wav -owts
source ./samples/gb0.wav.wts
ffplay ./samples/gb0.wav.mp4

https://user-images.githubusercontent.com/1991296/199337538-b7b0c7a3-2753-4a88-a0cd-f28a317987ba.mp4

---

Benchmarks

In order to have an objective comparison of the performance of the inference across different system configurations, use the bench tool. The tool simply runs the Encoder part of the model and prints how much time it took to execute it. The results are summarized in the following Github issue:

Benchmark results

ggml format

The original models are converted to a custom binary format. This allows to pack everything needed into a single file:

• model parameters
• mel filters
• vocabulary
• weights

You can download the converted models using the models/download-ggml-model.sh script or from here:

https://ggml.ggerganov.com

For more details, see the conversion script models/convert-pt-to-ggml.py or the README in models.

Bindings

• Rust: tazz4843/whisper-rs
• Objective-C / Swift: ggerganov/whisper.spm
• Python:
• Java:

Examples

There are various examples of using the library for different projects in the examples folder. Check them out!

Frequently asked questions (#126)