Merge pull request #21 from occ-ai/roy.bump_whisper_ver

chore: Update version to 0.0.9 and whispercpp to 1.6.2

buildspec.json

@@ -45,7 +45,7 @@
  }
  },
  "name": "obs-cleanstream",
- "version": "0.0.8",
+ "version": "0.0.9",
  "author": "Roy Shilkrot",
  "website": "https://github.com/occ-ai/obs-cleanstream/",
  "email": "[email protected]",

cmake/BuildWhispercpp.cmake

@@ -14,12 +14,12 @@ if(APPLE)
  endif(NOT DEFINED ENV{MACOS_ARCH})
 
  set(WHISPER_CPP_URL
- "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.2/whispercpp-macos-$ENV{MACOS_ARCH}-0.0.2.tar.gz"
+ "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.3/whispercpp-macos-$ENV{MACOS_ARCH}-0.0.3.tar.gz"
  )
  if($ENV{MACOS_ARCH} STREQUAL "x86_64")
- set(WHISPER_CPP_HASH "00C308AF0BFFF7619934403A8080CC9AFC4EDAA328D7587E617150A2C6A33313")
+ set(WHISPER_CPP_HASH "94DB35C69E958C8A59F0F331734F4F4F45F4BB13D2F54D3C838457E8590874C4")
  elseif($ENV{MACOS_ARCH} STREQUAL "arm64")
- set(WHISPER_CPP_HASH "0478E2079E07FA81BEE77506101003F4A4C8F0DF9E23757BD7E1D25DCBD1DB30")
+ set(WHISPER_CPP_HASH "ACA1DF8F34F4946B56FEED89B7548C9AD56D1DD89615C96BDEB6E4734A946451")
  else()
  message(
  FATAL_ERROR
@@ -59,16 +59,16 @@ elseif(WIN32)
  endif()
 
  set(WHISPER_CPP_URL
- "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.2/whispercpp-windows-${CUDA_PREFIX}-0.0.2.zip"
+ "https://github.com/occ-ai/occ-ai-dep-whispercpp/releases/download/0.0.3/whispercpp-windows-${CUDA_PREFIX}-0.0.3.zip"
  )
  if($ENV{CPU_OR_CUDA} STREQUAL "cpu")
- set(WHISPER_CPP_HASH "6DE628A51B9352624A1EC397231591FA3370E6BB42D9364F4F91F11DD18F77D2")
+ set(WHISPER_CPP_HASH "A7243E649E0B6D080AA6D2210DB0AC08C597FA11B88C3068B8A60083AD9E62EF")
  elseif($ENV{CPU_OR_CUDA} STREQUAL "clblast")
- set(WHISPER_CPP_HASH "97BF58520F1818B7C9F4E996197F3097934E5E0BBA92B0B016C6B28BE9FF1642")
+ set(WHISPER_CPP_HASH "7957AC76A0E6517C95951B3BECCB554CD992E30DAF8716681B40F375590F69F1")
  elseif($ENV{CPU_OR_CUDA} STREQUAL "12.2.0")
- set(WHISPER_CPP_HASH "48C059A3364E0AAD9FB0D4194BA554865928D22A27ECE5E3C116DC672D5D6EDE")
+ set(WHISPER_CPP_HASH "0F6BC1F91C573A867D6972554FC29C3D8EAFD7994FA0FEBBEAFCF945DC8A9F41")
  elseif($ENV{CPU_OR_CUDA} STREQUAL "11.8.0")
- set(WHISPER_CPP_HASH "29A5530E83896DE207F0199535CBBB24DF0D63B1373BA66139AD240BA67120EB")
+ set(WHISPER_CPP_HASH "51CB6750ADDF96F38106E4E88212FCC06500999E568E5A4EDC6D42CA6D7CA99D")
  else()
  message(
  FATAL_ERROR
@@ -106,11 +106,11 @@ elseif(WIN32)
  install(FILES ${WHISPER_DLLS} DESTINATION "obs-plugins/64bit")
 
 else()
- set(Whispercpp_Build_GIT_TAG "v1.5.5")
+ set(Whispercpp_Build_GIT_TAG "v1.6.2")
  set(WHISPER_EXTRA_CXX_FLAGS "-fPIC")
  set(WHISPER_ADDITIONAL_CMAKE_ARGS -DWHISPER_BLAS=OFF -DWHISPER_CUBLAS=OFF -DWHISPER_OPENBLAS=OFF)
 
- # On Linux and MacOS build a static Whisper library
+ # On Linux build a static Whisper library
  ExternalProject_Add(
  Whispercpp_Build
  DOWNLOAD_EXTRACT_TIMESTAMP true
@@ -131,7 +131,7 @@ else()
 
  ExternalProject_Get_Property(Whispercpp_Build INSTALL_DIR)
 
- # on Linux and MacOS add the static Whisper library to the link line
+ # add the static Whisper library to the link line
  add_library(Whispercpp::Whisper STATIC IMPORTED)
  set_target_properties(
  Whispercpp::Whisper

src/whisper-utils/whisper-processing.cpp

@@ -66,6 +66,7 @@ struct whisper_context *init_whisper_context(const std::string &model_path_in,
  cparams.use_gpu = false;
  obs_log(LOG_INFO, "Using CPU for inference");
 #endif
+ cparams.flash_attn = false;
 
  struct whisper_context *ctx = nullptr;
  try {
@@ -136,77 +137,78 @@ int run_whisper_inference(struct cleanstream_data *gf, const float *pcm32f_data,
  int(pcm32f_size), float(pcm32f_size) / WHISPER_SAMPLE_RATE,
  gf->whisper_params.n_threads);
 
+ std::string text_preproc;
+ float sentence_p = 0.0f;
+ int64_t t0;
+ int64_t t1;
+
  // run the inference
- int whisper_full_result = -1;
  try {
  gf->whisper_params.duration_ms =
  (int)((float)pcm32f_size / WHISPER_SAMPLE_RATE * 1000.0f);
- whisper_full_result = whisper_full(gf->whisper_context, gf->whisper_params,
- pcm32f_data, (int)pcm32f_size);
- } catch (const std::exception &e) {
- obs_log(LOG_ERROR, "Whisper exception: %s. Filter restart is required", e.what());
- whisper_free(gf->whisper_context);
- gf->whisper_context = nullptr;
- return DETECTION_RESULT_UNKNOWN;
- }
+ int whisper_full_result = whisper_full(gf->whisper_context, gf->whisper_params,
+ pcm32f_data, (int)pcm32f_size);
 
- if (whisper_full_result != 0) {
- obs_log(LOG_WARNING, "failed to process audio, error %d", whisper_full_result);
- return DETECTION_RESULT_UNKNOWN;
- } else {
+ if (whisper_full_result != 0) {
+ obs_log(LOG_WARNING, "failed to process audio, error %d",
+ whisper_full_result);
+ return DETECTION_RESULT_UNKNOWN;
+ }
+ if (whisper_full_n_segments(gf->whisper_context) == 0) {
+ return DETECTION_RESULT_SILENCE;
+ }
  const int n_segment = 0;
- const char *text = whisper_full_get_segment_text(gf->whisper_context, n_segment);
- const int64_t t0 = whisper_full_get_segment_t0(gf->whisper_context, n_segment);
- const int64_t t1 = whisper_full_get_segment_t1(gf->whisper_context, n_segment);
+ text_preproc = whisper_full_get_segment_text(gf->whisper_context, n_segment);
+ t0 = whisper_full_get_segment_t0(gf->whisper_context, n_segment);
+ t1 = whisper_full_get_segment_t1(gf->whisper_context, n_segment);
 
- float sentence_p = 0.0f;
  const int n_tokens = whisper_full_n_tokens(gf->whisper_context, n_segment);
  for (int j = 0; j < n_tokens; ++j) {
  sentence_p += whisper_full_get_token_p(gf->whisper_context, n_segment, j);
  }
  sentence_p /= (float)n_tokens;
+ } catch (const std::exception &e) {
+ obs_log(LOG_ERROR, "Whisper exception: %s. Filter restart is required", e.what());
+ whisper_free(gf->whisper_context);
+ gf->whisper_context = nullptr;
+ return DETECTION_RESULT_UNKNOWN;
+ }
 
- std::string text_preproc = text;
-
- if (text_preproc.empty()) {
- return DETECTION_RESULT_SILENCE;
- }
+ if (text_preproc.empty()) {
+ return DETECTION_RESULT_SILENCE;
+ }
 
- // if language is en convert text to lowercase
- if (strcmp(gf->whisper_params.language, "en") == 0) {
- std::string text_lower;
- std::transform(text_preproc.begin(), text_preproc.end(), text_lower.begin(),
- ::tolower);
- text_preproc = text_lower;
- // remove leading and trailing non-alphanumeric characters
- text_preproc = remove_leading_trailing_nonalpha(text_preproc);
- } else {
- // fix UTF8 encoding
- std::string text_fixed = fix_utf8(text);
- text_preproc = text_fixed;
- }
+ // if language is en convert text to lowercase
+ if (strcmp(gf->whisper_params.language, "en") == 0) {
+ std::transform(text_preproc.begin(), text_preproc.end(), text_preproc.begin(),
+ ::tolower);
+ // remove leading and trailing non-alphanumeric characters
+ text_preproc = remove_leading_trailing_nonalpha(text_preproc);
+ } else {
+ // fix UTF8 encoding
+ text_preproc = fix_utf8(text_preproc);
+ }
 
-  if (gf->log_words) {
-  obs_log(LOG_INFO, "[%s --> %s] (%.3f) %s", to_timestamp(t0).c_str(),
-  to_timestamp(t1).c_str(), sentence_p, text_preproc.c_str());
-  }
+ if (gf->log_words) {
+ obs_log(LOG_INFO, "[%s --> %s] (%.3f) %s", to_timestamp(t0).c_str(),
+ to_timestamp(t1).c_str(), sentence_p, text_preproc.c_str());
+ }
 
-  if (text_preproc.empty()) {
-  return DETECTION_RESULT_SILENCE;
-  }
+ if (text_preproc.empty()) {
+ return DETECTION_RESULT_SILENCE;
+ }
 
- // use a regular expression to detect filler words with a word boundary
- try {
- if (gf->detect_regex != nullptr && strlen(gf->detect_regex) > 0) {
- std::regex filler_regex(gf->detect_regex);
- if (std::regex_search(text_preproc, filler_regex,
- std::regex_constants::match_any)) {
- return DETECTION_RESULT_BEEP;
- }
+ // use a regular expression to detect filler words with a word boundary
+ try {
+ if (gf->detect_regex != nullptr && strlen(gf->detect_regex) > 0) {
+ std::regex filler_regex(gf->detect_regex);
+ if (std::regex_search(text_preproc, filler_regex,
+ std::regex_constants::match_any)) {
+ return DETECTION_RESULT_BEEP;
  }
- } catch (const std::regex_error &e) {
- obs_log(LOG_ERROR, "Regex error: %s", e.what());
  }
+ } catch (const std::regex_error &e) {
+ obs_log(LOG_ERROR, "Regex error: %s", e.what());
  }
 
  return DETECTION_RESULT_SPEECH;