{"id":19016,"date":"2026-09-12T13:03:41","date_gmt":"2026-09-12T11:03:41","guid":{"rendered":"https:\/\/www.beseit.net\/?p=19016"},"modified":"2026-09-12T17:18:59","modified_gmt":"2026-09-12T15:18:59","slug":"integracio-de-speech-dispatcher-amb-piper-tts-veu-ona-upc-a-deepin","status":"publish","type":"post","link":"https:\/\/www.beseit.net\/?p=19016","title":{"rendered":"Integraci\u00f3 de Speech Dispatcher amb Piper TTS (veu Ona, UPC) a Deepin"},"content":{"rendered":"\n

Data: 12 de setembre de 2026 \u2014 Estat actual del projecte<\/em><\/h1>\n\n\n\n

\ud83c\udfaf Objectiu<\/h2>\n\n\n\n

Aconseguir que Firefox pugui llegir p\u00e0gines web en catal\u00e0<\/strong> amb la veu Ona (UPC)<\/strong> mitjan\u00e7ant la cadena:<\/p>\n\n\n\n

Firefox \u2192 Speech Dispatcher \u2192 m\u00f2dul gen\u00e8ric (sd_generic) \u2192 script pont \u2192 Piper (Docker) \u2192 Veu Ona<\/pre>\n\n\n\n
El repte: Speech Dispatcher no detecta cap veu per defecte (0 veus), perqu\u00e8 Piper no \u00e9s un motor TTS natiu de Speech Dispatcher.<\/p>\n\n\n\n
\n\n\n\n
\ud83c\udfd7\ufe0f Arquitectura de la soluci\u00f3<\/h2>\n\n\n\n
Component<\/th>Paper<\/th><\/tr><\/thead>
Firefox<\/strong><\/td>Client que demana s\u00edntesi de veu a Speech Dispatcher<\/td><\/tr>
Speech Dispatcher<\/strong><\/td>Daemon intermediari (0.11.5)<\/td><\/tr>
sd_generic<\/code><\/strong><\/td>M\u00f2dul gen\u00e8ric de Speech Dispatcher que executa ordres externes<\/td><\/tr>
piper-spd<\/code><\/strong><\/td>Script pont: rep text, truca a Docker, retorna WAV<\/td><\/tr>
Piper (Docker)<\/strong><\/td>Motor TTS al port 10200 (protocol Wyoming)<\/td><\/tr>
Veu ca_ES-upc_ona-medium<\/code><\/strong><\/td>Veu catalana d’Ona, desenvolupada a la UPC<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\n\n\n\n
\ud83d\udcc2 Fitxers creats<\/h2>\n\n\n\n
1. Script pont \/usr\/local\/bin\/piper-spd<\/code><\/h3>\n\n\n\n
Rep el text de Speech Dispatcher, el passa al contenidor Piper i genera el fitxer WAV:<\/p>\n\n\n\n
#!\/bin\/bash\n# piper-spd: pont entre Speech Dispatcher i Piper TTS (veus UPC)\nMODEL=\"\/config\/ca_ES-upc_ona-medium.onnx\"\nOUTPUT=\"\"\nTEXT=\"\"\nVOICE=\"${VOICE:-ona}\"\n\nwhile [ $# -gt 0 ]; do\n  case \"$1\" in\n    -o) OUTPUT=\"$2\"; shift 2;;\n    -l|--list)\n      echo \"ona ca_ES medium\"\n      echo \"pau ca_ES medium\"\n      exit 0;;\n    -*) shift;;\n    *) TEXT=\"$TEXT $1\"; shift;;\n  esac\ndone\n\n[ -z \"$OUTPUT\" ] && { echo \"Error: falta -o FITXER\" >&2; exit 1; }\n\ncase \"$VOICE\" in\n  pau) MODEL=\"\/config\/ca_ES-upc_pau-medium.onnx\";;\n  *)   MODEL=\"\/config\/ca_ES-upc_ona-medium.onnx\";;\nesac\n\nif ! docker ps --format '{{.Names}}' | grep -q '^piper$'; then\n  echo \"Error: el contenidor piper no est\u00e0 en marxa\" >&2\n  exit 1\nfi\n\ndocker exec piper bash -c \"printf '%s' \\\"$TEXT\\\" | piper --model $MODEL --output_file \/tmp\/piper_spd.wav\" 2>\/dev\/null\nif ! docker cp piper:\/tmp\/piper_spd.wav \"$OUTPUT\" 2>\/dev\/null; then\n  echo \"Error: no s'ha pogut generar l'\u00e0udio\" >&2\n  exit 1\nfi\ndocker exec piper rm -f \/tmp\/piper_spd.wav 2>\/dev\/null\nexit 0<\/pre>\n\n\n\n
2. Configuraci\u00f3 del m\u00f2dul \/etc\/speech-dispatcher\/modules\/piper-generic.conf<\/code><\/h3>\n\n\n\n
GenericExecuteSynth \"VOICE=$VOICE \/usr\/local\/bin\/piper-spd -o $FILE $TEXT\"\nGenericSsmlMode \"off\"\nGenericLanguage \"ca\" \"ca\"\n\nAddVoice \"ca\" \"FEMALE1\" \"ona\"\ndefaultVoice \"ona\"<\/pre>\n\n\n\n
3. Registre a \/etc\/speech-dispatcher\/speechd.conf<\/code><\/h3>\n\n\n\n
AddModule \"piper\" \"sd_generic\" \"\/etc\/speech-dispatcher\/modules\/piper-generic.conf\"<\/pre>\n\n\n\n
\n\n\n\n
\ud83e\uddd7 Obstacles superats (i com es van resoldre)<\/h2>\n\n\n\n
    \n
  1. M\u00f2dul sd_generic<\/code> no localitzable<\/strong> \u2192 Resulta que es troba a \/usr\/lib\/speech-dispatcher-modules\/<\/code> (sense subdirectori intermedi).<\/li>\n\n\n\n
  2. Speech Dispatcher es negava a arrencar<\/strong> \u2192 El directori de depuraci\u00f3 \/tmp\/speechd-debug<\/code> bloquejava l’inici; cal esborrar-lo abans d’arrencar.<\/li>\n\n\n\n
  3. Error --voice is required for the piper backend<\/code><\/strong> (Docker) \u2192 Calia especificar PIPER_VOICE=ca_ES-upc_ona-medium<\/code> en crear el contenidor.<\/li>\n\n\n\n
  4. aplay<\/code> dins del contenidor no funcionava<\/strong> \u2192 La s\u00edntesi es fa al contenidor i la reproducci\u00f3 a l’amfitri\u00f3.<\/li>\n\n\n\n
  5. Error \u00abThe module does not have any voice configured\u00bb<\/em><\/strong> \u2192 Faltava la secci\u00f3 AddVoice<\/code><\/strong> a la configuraci\u00f3 del m\u00f2dul gen\u00e8ric. Aquest era el pas definitiu! \u2705<\/li>\n<\/ol>\n\n\n\n
    \n\n\n\n
    \u2705 Estat actual<\/h2>\n\n\n\n