Add --verbose flag for colored logging

CMakeLists.txt

@@ -18,6 +18,16 @@ endif()
 # Add path to modules
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/")
 
+# Add termcolor for colored terminal output
+include(FetchContent)
+FetchContent_Declare(
+    termcolor
+    GIT_REPOSITORY https://github.com/ikalnytskyi/termcolor.git
+    GIT_TAG        v2.1.0
+    GIT_SHALLOW    TRUE
+)
+FetchContent_MakeAvailable(termcolor)
+
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
 
 if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)

examples/cli/CMakeLists.txt

@@ -3,6 +3,6 @@ add_executable(${TARGET} cli.cpp)
 
 include(DefaultTargetOptions)
 
-target_link_libraries(${TARGET} PRIVATE common whisper ${FFMPEG_LIBRARIES} ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE common whisper termcolor::termcolor ${FFMPEG_LIBRARIES} ${CMAKE_THREAD_LIBS_INIT})
 
 install(TARGETS ${TARGET} RUNTIME)

examples/cli/cli.cpp

@@ -5,14 +5,18 @@
 #include "grammar-parser.h"
 
 #include <cmath>
+#include <cstdint>
 #include <fstream>
 #include <cstdio>
+#include <iostream>
 #include <string>
 #include <thread>
 #include <vector>
 #include <cstring>
 #include <cfloat>
 
+#include <termcolor/termcolor.hpp>
+
 #if defined(_WIN32)
 #ifndef NOMINMAX
 #define NOMINMAX
@@ -77,6 +81,7 @@ struct whisper_params {
     bool use_gpu         = true;
     bool flash_attn      = true;
     bool suppress_nst    = false;
+    bool verbose         = false;
 
     std::string language  = "en";
     std::string prompt;
@@ -208,6 +213,7 @@ static bool whisper_params_parse(int argc, char ** argv, whisper_params & params
         else if (arg == "-vmsd" || arg == "--vad-max-speech-duration-s")   { params.vad_max_speech_duration_s   = std::stof(ARGV_NEXT); }
         else if (arg == "-vp"   || arg == "--vad-speech-pad-ms")           { params.vad_speech_pad_ms           = std::stoi(ARGV_NEXT); }
         else if (arg == "-vo"   || arg == "--vad-samples-overlap")         { params.vad_samples_overlap         = std::stof(ARGV_NEXT); }
+        else if (arg == "-v"    || arg == "--verbose")                     { params.verbose                     = true; }
         else {
             fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
             whisper_print_usage(argc, argv, params);
@@ -258,6 +264,7 @@ static void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params
     fprintf(stderr, "  -ojf,      --output-json-full  [%-7s] include more information in the JSON file\n",      params.output_jsn_full ? "true" : "false");
     fprintf(stderr, "  -of FNAME, --output-file FNAME [%-7s] output file path (without file extension)\n",      "");
     fprintf(stderr, "  -np,       --no-prints         [%-7s] do not print anything other than the results\n",   params.no_prints ? "true" : "false");
+    fprintf(stderr, "  -v,        --verbose           [%-7s] enable verbose output (show INFO level messages)\n",  params.verbose ? "true" : "false");
     fprintf(stderr, "  -ps,       --print-special     [%-7s] print special tokens\n",                           params.print_special ? "true" : "false");
     fprintf(stderr, "  -pc,       --print-colors      [%-7s] print colors\n",                                   params.print_colors ? "true" : "false");
     fprintf(stderr, "             --print-confidence  [%-7s] print confidence\n",                               params.print_confidence ? "true" : "false");
@@ -910,6 +917,41 @@ static void output_lrc(struct whisper_context * ctx, std::ofstream & fout, const
 
 static void cb_log_disable(enum ggml_log_level , const char * , void * ) { }
 
+// Custom log callback that filters INFO messages based on verbose flag
+struct log_filter_data {
+    bool verbose;
+};
+
+static void cb_log_filter(enum ggml_log_level level, const char * text, void * user_data) {
+    log_filter_data * data = (log_filter_data *) user_data;
+
+    // Apply colors based on log level (same as whisper.cpp default callback)
+    switch (level) {
+        case GGML_LOG_LEVEL_ERROR:
+            std::cerr << termcolor::red << text << termcolor::reset;
+            break;
+        case GGML_LOG_LEVEL_WARN:
+            std::cerr << termcolor::yellow << text << termcolor::reset;
+            break;
+        case GGML_LOG_LEVEL_INFO:
+            // Show info messages only if verbose is enabled
+            if (data->verbose) {
+                std::cerr << termcolor::cyan << text << termcolor::reset;
+            }
+            break;
+        case GGML_LOG_LEVEL_DEBUG:
+            // Show debug messages only in debug mode
+            #ifdef WHISPER_DEBUG
+            std::cerr << text;
+            #endif
+            break;
+        default:
+            std::cerr << text;
+            break;
+    }
+    std::cerr.flush();
+}
+
 int main(int argc, char ** argv) {
     ggml_backend_load_all();
 
@@ -987,8 +1029,14 @@ int main(int argc, char ** argv) {
         exit(0);
     }
 
+    // Setup logging based on flags
+    static log_filter_data log_data;
+    log_data.verbose = params.verbose;
+
     if (params.no_prints) {
         whisper_log_set(cb_log_disable, NULL);
+    } else {
+        whisper_log_set(cb_log_filter, &log_data);
     }
 
     // whisper init

src/CMakeLists.txt

@@ -126,6 +126,7 @@ if (WHISPER_EXTRA_FLAGS)
 endif()
 
 target_link_libraries(whisper PUBLIC ggml)
+target_link_libraries(whisper PRIVATE termcolor::termcolor)
 
 if (WHISPER_COREML)
     target_link_libraries(whisper PRIVATE whisper.coreml)

src/whisper.cpp

@@ -14,6 +14,8 @@
 #include "openvino/whisper-openvino-encoder.h"
 #endif
 
+#include <termcolor/termcolor.hpp>
+
 #include <atomic>
 #include <algorithm>
 #include <cassert>
@@ -22,11 +24,13 @@
 #include <cmath>
 #include <climits>
 #include <codecvt>
+#include <cstdint>
 #include <cstdarg>
 #include <cstdio>
 #include <cstring>
 #include <fstream>
 #include <functional>
+#include <iostream>
 #include <map>
 #include <mutex>
 #include <random>
@@ -1841,6 +1845,10 @@ static bool whisper_model_load(struct whisper_model_loader * loader, whisper_con
 
             size_t size_main = ggml_backend_buffer_get_size(buf);
             WHISPER_LOG_INFO("%s: %12s total size = %8.2f MB\n", __func__, ggml_backend_buffer_name(buf), size_main / 1e6);
+        } else {
+            WHISPER_LOG_ERROR("%s: failed to allocate backend buffer for %s\n", __func__, ggml_backend_buft_name(buft));
+            WHISPER_LOG_ERROR("%s: not enough memory available - try using a smaller model or reducing GPU usage\n", __func__);
+            return false;
         }
     }
 
@@ -4970,14 +4978,16 @@ struct whisper_vad_context * whisper_vad_init_with_params(
 
             size_t size_main = ggml_backend_buffer_get_size(buf);
             WHISPER_LOG_INFO("%s: %12s total size = %8.2f MB\n", __func__, ggml_backend_buffer_name(buf), size_main / 1e6);
+        } else {
+            WHISPER_LOG_ERROR("%s: failed to allocate backend buffer for %s\n", __func__, ggml_backend_buft_name(buft));
+            WHISPER_LOG_ERROR("%s: not enough memory available - try using a smaller model or reducing GPU usage\n", __func__);
+            return nullptr;
         }
     }
 
     // load weights
     {
         size_t total_size = 0;
-        model.n_loaded = 0;
-        std::vector<char> read_buf;
 
         while (true) {
             int32_t n_dims;
@@ -8938,13 +8948,27 @@ static void whisper_log_internal(ggml_log_level level, const char * format, ...)
 }
 
 static void whisper_log_callback_default(ggml_log_level level, const char * text, void * user_data) {
-    (void) level;
     (void) user_data;
 #ifndef WHISPER_DEBUG
     if (level == GGML_LOG_LEVEL_DEBUG) {
         return;
     }
 #endif
-    fputs(text, stderr);
+
+    // Apply colors based on log level
+    switch (level) {
+        case GGML_LOG_LEVEL_ERROR:
+            std::cerr << termcolor::red << text << termcolor::reset;
+            break;
+        case GGML_LOG_LEVEL_WARN:
+            std::cerr << termcolor::yellow << text << termcolor::reset;
+            break;
+        case GGML_LOG_LEVEL_INFO:
+            std::cerr << termcolor::cyan << text << termcolor::reset;
+            break;
+        default:
+            fputs(text, stderr);
+            break;
+    }
     fflush(stderr);
 }