o Wli%@sdZddlZddlZddlZddlZddlZddlmZddlm Z m Z m Z m Z ddl mZddlmZmZddlmZmZdd lmZmZeZdd d Zed krXedSdS)z)Flask web server with HTTP API for Piper.N)Path)AnyDictListOptional)urlopen)Flaskrequest) PiperVoiceSynthesisConfig) VOICES_JSONdownload_voicereturnc szt}|jdddd|jdtddd|jd d d d d |jddtdd|jddtdd|jddtdd|jddddtdd|jdddd |jd!d"td#d$d|jd%d&d'ttgd(d)|jd*d+d,d-|jd.dd/d |t j j rt j nt j d0t jsjd1_tjtjsˆj}jD]}t||d2}t d3||r|nqstd4d5jd2tjjd6itt}|jd7d8gd9d:tttfffd;d< }|jd=d8gd9d:tttffd>d?}|jd@dAgd9d:tffdBdC }|jdDdAgd9d:tffdEdF }|jj j!dGdHS)IzRun HTTP server.z--hostz0.0.0.0zHTTP server host)defaulthelpz--portizHTTP server port)typerrz-mz--modelTzPath to Onnx model file)requiredrz-sz --speakerzId of speaker (default: 0))rrz--length-scalez--length_scalezPhoneme lengthz --noise-scalez --noise_scalezGenerator noisez--noise-w-scalez--noise_w_scalez --noise-wz --noise_wzPhoneme width noisez--cuda store_truezUse GPU)actionrz--sentence-silencez--sentence_silencegz&Seconds of silence after each sentencez --data-dirz --data_dirappendzJData directory to check for downloaded models (default: current directory))rrrz--download-dirz--download_dirz1Path to download voices (default: first data dir))rz--debugzPrint DEBUG messages to console)levelr.onnxz Checking '%s'zUnable to find voice: z (use piper.download_voices)Zuse_cudaz/voicesGET)methodsrc si}tdg}jD]}t|dD]}t|d}|r(||qq |D]+}|jd}||vr9q,t|ddd}t |||<Wdn1sRwYq,|S)zList downloaded voices. Outputs a JSON object with the format: { "": { }, ... } for each voice in your data directories. z.jsonz*.onnxz .onnx.jsonrzutf-8)encodingN) rdata_dirglobexistsrnamerstripopenjsonload)Z voices_dictZ config_pathsrZ onnx_pathZ config_pathmodel_id config_file)args model_pathK/home/kim/smarthome/.venv/lib/python3.10/site-packages/piper/http_server.py app_voicesjs$    zmain..app_voicesz /all-voicescSs6tt }t|WdS1swYdS)zList all Piper voices. Outputs voices.json from the piper-voices repo on HuggingFace. See: https://huggingface.co/rhasspy/piper-voices N)rr r$r%)responser*r*r+app_all_voicess $zmain..app_all_voicesz /downloadPOSTcs@ttj}|d}|std|dd}t||d|S)aDownload a voice. Downloads the .onnx and .onnx.json file from piper-voices repo on HuggingFace. See: https://huggingface.co/rhasspy/piper-voices Expects a JSON object with the format: { "voice": "", (required) "force_redownload": false (optional) } Returns the name of the voice. Voice format must be -- like "en_US-lessac-medium". voicezvoice is requiredforce_redownloadF)r1)r$loadsr dataget ValueErrorr)r3r&r1) download_dirr*r+ app_downloads   zmain..app_download/c sttj}|dd}|stdt||d}|}|durOj D]#}t ||d}| rNtd|t j |jd}||<nq+|dur[td |}|d }|jjd kr|dur|d }|rx|jj|}|durtd ||jjjpd}|dur||jjkrd}t|t|djdurjn|jjt|djdurjn|jjt|djdurȈjn|jjd}td||tf} t| d} | Id} t|||D]6\} } | s | | j!| "| j#| $| j%d} | dkr| &t't(|jj!j)d| &| j*qWdn 1s1wY| +WdS1sEwYdS)aSynthesize audio from text. Expects a JSON object with the format: { "text": "Text to speak.", (required) "voice": "", (optional) "speaker": "", (optional) "speaker_id": "", (optional, overrides speaker) "length_scale": 1.0, (optional) "noise_scale": 0.667, (optional) "length_w_scale": 0.8 (optional) } textzNo text providedr0NrzLoading voice %srz)Voice not found: %s. Using default voice. speaker_idr speakerzSpeaker not found: '%s' in %sr length_scale noise_scale noise_w_scale)r;r=r>r?z&Synthesizing text: '%s' with config=%swbFT),r$r2r r3r4stripr5_LOGGERdebugrrr r r%cudawarningconfigZ num_speakersZspeaker_id_mapkeysr<r floatr=r>r?ioBytesIOwaver# enumerateZ synthesizeZ setframerateZ sample_rateZ setsampwidthZ sample_widthZ setnchannelsZsample_channelsZ writeframesbytesintZsentence_silenceZaudio_int16_bytesgetvalue)r3r9r&r0rmaybe_model_pathr;r<Z syn_configZwav_ioZwav_fileZwav_params_setiZ audio_chunk)r(default_model_id default_voice loaded_voicesr*r+app_synthesizes                "      &zmain..app_synthesize)hostportN)"argparseArgumentParser add_argumentrOrIstrrcwd parse_argslogging basicConfigrDDEBUGINFOrCr6rmodelr r5r!r"r r%rEr__name__ZrouterrrNrunrWrX) parserZ voice_namerrQZappr,r.r7rVr*)r(rSrTr6rUr)r+mains           prg__main__)rN)__doc__rYrJr$r_rLpathlibrtypingrrrrurllib.requestrZflaskrr r:r r Zdownload_voicesr r getLoggerrCrgrdr*r*r*r+s&