Hello all. I am able to programmatically decode speex just fine (playing others' encodings), but my encoding eats the beginning of words. If I encode a word that gradually increases in volume, like "wonderful", I hear "nderful", but if I encode something percussive like "beep" I hear almost all of it. It's as if the modeller does not detect the start of a word. Has anyone seen this before? In case you're interested here is the init code: int quality = 3; int off = 0; speex_bits_init(&encoderBits); encoder = speex_encoder_init(&speex_nb_mode); speex_encoder_ctl( encoder, SPEEX_GET_SAMPLING_RATE, &sampleRate ); speex_encoder_ctl( encoder, SPEEX_GET_FRAME_SIZE, &frameSize); preprocess = speex_preprocess_state_init( frameSize, sampleRate ); speex_preprocess_ctl( preprocess, SPEEX_SET_QUALITY, (void*)&quality); speexRecordData = (char*)malloc(frameSize*sizeof(char)*CHANNELS); // 2 channels, input and output and the encoding code: speex_bits_reset( &encoderBits ); if ( speex_preprocess( preprocess, (short*)recording.data(), 0 ) && speex_encode_int( encoder, (short*)recording.data(), &encoderBits ) ) // encode to speex { int numBytes = speex_bits_write( &encoderBits, speexRecordData, frameSize ); ... Thanks in advance for your insight. yx -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20070912/08b7deb5/attachment.html