whisperfile server: convert files without ffmpeg (#568)

Author: cjpais

whisper.cpp/common.cpp

@@ -37,21 +37,6 @@ static void on_exit(void) {
     }
 }
 
-bool is_wav_buffer(const std::string buf) {
-    // RIFF ref: https://en.wikipedia.org/wiki/Resource_Interchange_File_Format
-    // WAV ref: https://www.mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/WAVE.html
-    if (buf.size() < 12 || buf.substr(0, 4) != "RIFF" || buf.substr(8, 4) != "WAVE") {
-        return false;
-    }
-
-    uint32_t chunk_size = *reinterpret_cast<const uint32_t*>(buf.data() + 4);
-    if (chunk_size + 8 != buf.size()) {
-        return false;
-    }
-
-    return true;
-}
-
 static ma_result perform_audio_conversion(ma_decoder* pDecoder, ma_encoder* pEncoder) {
     ma_result rc = MA_SUCCESS;
     for (;;) {
@@ -173,12 +158,6 @@ bool read_wav(const std::string & fname_, std::vector<float>& pcmf32, std::vecto
 
         fprintf(stderr, "%s: read %zu bytes from stdin\n", __func__, wav_data.size());
     }
-    else if (is_wav_buffer(fname)) {
-        if (drwav_init_memory(&wav, fname.c_str(), fname.size(), nullptr) == false) {
-            fprintf(stderr, "error: failed to open WAV file from fname buffer\n");
-            return false;
-        }
-    }
     else if (drwav_init_file(&wav, fname.c_str(), nullptr) == false) {
         tinylogf("%s: converting to wav...\n", fname.c_str());
         TRY_CONVERSION;

whisper.cpp/server.cpp

@@ -732,13 +732,16 @@ int whisper_server_main(int argc, char ** argv) {
         std::vector<float> pcmf32;               // mono-channel F32 PCM
         std::vector<std::vector<float>> pcmf32s; // stereo-channel F32 PCM
 
+        // write incoming buffer to temporary file
+        std::string temp_filename = __get_tmpdir();
+        temp_filename += "/whisperfile.";
+        temp_filename += std::to_string(_rand64());
+
+        std::ofstream temp_file{temp_filename, std::ios::binary};
+        temp_file << audio_file.content;
+        temp_file.close();
+
         if (sparams.ffmpeg_converter) {
-            // if file is not wav, convert to wav
-            // write to temporary file
-            const std::string temp_filename = "whisper_server_temp_file.wav";
-            std::ofstream temp_file{temp_filename, std::ios::binary};
-            temp_file << audio_file.content;
-            temp_file.close();
 
             std::string error_resp = "{\"error\":\"Failed to execute ffmpeg command.\"}";
             const bool is_converted = convert_to_wav(temp_filename, error_resp);
@@ -756,17 +759,17 @@ int whisper_server_main(int argc, char ** argv) {
                 std::remove(temp_filename.c_str());
                 return;
             }
-            // remove temp file
-            std::remove(temp_filename.c_str());
         } else {
-            if (!::read_wav(audio_file.content, pcmf32, pcmf32s, params.diarize))
+            if (!::read_wav(temp_filename, pcmf32, pcmf32s, params.diarize))
             {
                 fprintf(stderr, "error: failed to read WAV file\n");
                 const std::string error_resp = "{\"error\":\"failed to read WAV file\"}";
                 res.set_content(error_resp, "application/json");
                 return;
             }
         }
+        // remove temp file
+        std::remove(temp_filename.c_str());
 
 
         printf("Successfully loaded %s\n", filename.c_str());