Skip to content

server : add Voice Activity Detection (VAD) support #3246

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Jun 13, 2025
+91 −0
Merged

server : add Voice Activity Detection (VAD) support #3246

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
63fb077
server : add Voice Activity Detection (VAD) support
danbev Jun 13, 2025
9c829d2
server : add VAD parameters to usage in README.md [no ci]
danbev Jun 13, 2025
b8658ad
server : fix conflicting short options [no ci]
danbev Jun 13, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 20 additions & 0 deletions examples/server/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ options:
-sow, --split-on-word [false ] split on word rather than on token
-bo N, --best-of N [2 ] number of best candidates to keep
-bs N, --beam-size N [-1 ] beam size for beam search
-ac N, --audio-ctx N [0 ] audio context size (0 - all)
-wt N, --word-thold N [0.01 ] word timestamp probability threshold
-et N, --entropy-thold N [2.40 ] entropy threshold for decoder fail
-lpt N, --logprob-thold N [-1.00 ] log probability threshold for decoder fail
Expand All @@ -41,9 +42,28 @@ options:
--prompt PROMPT [ ] initial prompt
-m FNAME, --model FNAME [models/ggml-base.en.bin] model path
-oved D, --ov-e-device DNAME [CPU ] the OpenVINO device used for encode inference
-dtw MODEL --dtw MODEL [ ] compute token-level timestamps
--host HOST, [127.0.0.1] Hostname/ip-adress for the server
--port PORT, [8080 ] Port number for the server
--public PATH, [examples/server/public] Path to the public folder
--request-path PATH, [ ] Request path for all requests
--inference-path PATH, [/inference] Inference path for all requests
--convert, [false ] Convert audio to WAV, requires ffmpeg on the server
-sns, --suppress-nst [false ] suppress non-speech tokens
-nth N, --no-speech-thold N [0.60 ] no speech threshold
-nc, --no-context [false ] do not use previous audio context
-ng, --no-gpu [false ] do not use gpu
-fa, --flash-attn [false ] flash attention

Voice Activity Detection (VAD) options:
--vad [false ] enable Voice Activity Detection (VAD)
-vm FNAME, --vad-model FNAME [ ] VAD model path
-vt N, --vad-threshold N [0.50 ] VAD threshold for speech recognition
-vspd N, --vad-min-speech-duration-ms N [250 ] VAD min speech duration (0.0-1.0)
-vsd N, --vad-min-silence-duration-ms N [100 ] VAD min silence duration (to split segments)
-vmsd N, --vad-max-speech-duration-s N [FLT_MAX] VAD max speech duration (auto-split longer)
-vp N, --vad-speech-pad-ms N [30 ] VAD speech padding (extend segments)
-vo N, --vad-samples-overlap N [0.10 ] VAD samples overlap (seconds between segments)
```

> [!WARNING]
Expand Down
71 changes: 71 additions & 0 deletions examples/server/server.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
#include "httplib.h"
#include "json.hpp"

#include <cfloat>
#include <chrono>
#include <cmath>
#include <cstdio>
Expand Down Expand Up @@ -90,6 +91,16 @@ struct whisper_params {
std::string openvino_encode_device = "CPU";

std::string dtw = "";

// Voice Activity Detection (VAD) parameters
bool vad = false;
std::string vad_model = "";
float vad_threshold = 0.5f;
int vad_min_speech_duration_ms = 250;
int vad_min_silence_duration_ms = 100;
float vad_max_speech_duration_s = FLT_MAX;
int vad_speech_pad_ms = 30;
float vad_samples_overlap = 0.1f;
};

void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & params, const server_params& sparams) {
Expand Down Expand Up @@ -140,6 +151,18 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
fprintf(stderr, " -nc, --no-context [%-7s] do not use previous audio context\n", params.no_context ? "true" : "false");
fprintf(stderr, " -ng, --no-gpu [%-7s] do not use gpu\n", params.use_gpu ? "false" : "true");
fprintf(stderr, " -fa, --flash-attn [%-7s] flash attention\n", params.flash_attn ? "true" : "false");
// Voice Activity Detection (VAD) parameters
fprintf(stderr, "\nVoice Activity Detection (VAD) options:\n");
fprintf(stderr, " --vad [%-7s] enable Voice Activity Detection (VAD)\n", params.vad ? "true" : "false");
fprintf(stderr, " -vm FNAME, --vad-model FNAME [%-7s] VAD model path\n", params.vad_model.c_str());
fprintf(stderr, " -vt N, --vad-threshold N [%-7.2f] VAD threshold for speech recognition\n", params.vad_threshold);
fprintf(stderr, " -vspd N, --vad-min-speech-duration-ms N [%-7d] VAD min speech duration (0.0-1.0)\n", params.vad_min_speech_duration_ms);
fprintf(stderr, " -vsd N, --vad-min-silence-duration-ms N [%-7d] VAD min silence duration (to split segments)\n", params.vad_min_silence_duration_ms);
fprintf(stderr, " -vmsd N, --vad-max-speech-duration-s N [%-7s] VAD max speech duration (auto-split longer)\n", params.vad_max_speech_duration_s == FLT_MAX ?
std::string("FLT_MAX").c_str() :
std::to_string(params.vad_max_speech_duration_s).c_str());
fprintf(stderr, " -vp N, --vad-speech-pad-ms N [%-7d] VAD speech padding (extend segments)\n", params.vad_speech_pad_ms);
fprintf(stderr, " -vo N, --vad-samples-overlap N [%-7.2f] VAD samples overlap (seconds between segments)\n", params.vad_samples_overlap);
fprintf(stderr, "\n");
}

Expand Down Expand Up @@ -195,6 +218,16 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params, serve
else if ( arg == "--request-path") { sparams.request_path = argv[++i]; }
else if ( arg == "--inference-path") { sparams.inference_path = argv[++i]; }
else if ( arg == "--convert") { sparams.ffmpeg_converter = true; }

// Voice Activity Detection (VAD)
else if ( arg == "--vad") { params.vad = true; }
else if (arg == "-vm" || arg == "--vad-model") { params.vad_model = argv[++i]; }
else if (arg == "-vt" || arg == "--vad-threshold") { params.vad_threshold = std::stof(argv[++i]); }
else if (arg == "-vspd" || arg == "--vad-min-speech-duration-ms") { params.vad_min_speech_duration_ms = std::stoi(argv[++i]); }
else if (arg == "-vsd" || arg == "--vad-min-silence-duration-ms") { params.vad_min_speech_duration_ms = std::stoi(argv[++i]); }
else if (arg == "-vmsd" || arg == "--vad-max-speech-duration-s") { params.vad_max_speech_duration_s = std::stof(argv[++i]); }
else if (arg == "-vp" || arg == "--vad-speech-pad-ms") { params.vad_speech_pad_ms = std::stoi(argv[++i]); }
else if (arg == "-vo" || arg == "--vad-samples-overlap") { params.vad_samples_overlap = std::stof(argv[++i]); }
else {
fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
whisper_print_usage(argc, argv, params, sparams);
Expand Down Expand Up @@ -511,6 +544,34 @@ void get_req_parameters(const Request & req, whisper_params & params)
{
params.no_context = parse_str_to_bool(req.get_file_value("no_context").content);
}
if (req.has_file("vad"))
{
params.vad = parse_str_to_bool(req.get_file_value("vad").content);
}
if (req.has_file("vad_threshold"))
{
params.vad_threshold = std::stof(req.get_file_value("vad_threshold").content);
}
if (req.has_file("vad_min_speech_duration_ms"))
{
params.vad_min_speech_duration_ms = std::stof(req.get_file_value("vad_min_speech_duration_ms").content);
}
if (req.has_file("vad_min_silence_duration_ms"))
{
params.vad_min_silence_duration_ms = std::stof(req.get_file_value("vad_min_silence_duration_ms").content);
}
if (req.has_file("vad_max_speech_duration_s"))
{
params.vad_max_speech_duration_s = std::stof(req.get_file_value("vad_max_speech_duration_s").content);
}
if (req.has_file("vad_speech_pad_ms"))
{
params.vad_speech_pad_ms = std::stoi(req.get_file_value("vad_speech_pad_ms").content);
}
if (req.has_file("vad_samples_overlap"))
{
params.vad_samples_overlap = std::stof(req.get_file_value("vad_samples_overlap").content);
}
}

} // namespace
Expand Down Expand Up @@ -829,6 +890,16 @@ int main(int argc, char ** argv) {

wparams.suppress_nst = params.suppress_nst;

wparams.vad = params.vad;
wparams.vad_model_path = params.vad_model.c_str();

wparams.vad_params.threshold = params.vad_threshold;
wparams.vad_params.min_speech_duration_ms = params.vad_min_speech_duration_ms;
wparams.vad_params.min_silence_duration_ms = params.vad_min_silence_duration_ms;
wparams.vad_params.max_speech_duration_s = params.vad_max_speech_duration_s;
wparams.vad_params.speech_pad_ms = params.vad_speech_pad_ms;
wparams.vad_params.samples_overlap = params.vad_samples_overlap;

whisper_print_user_data user_data = { &params, &pcmf32s, 0 };

// this callback is called on each new segment
Expand Down