Njikwa olu na PC na RaspberryPi na Whisper

Echiche nke oru ngo bụ nye ntuziaka olu ka ị na-emekọrịta ihe site na PC ma ọ bụ Raspberry Pi anyị site na iji ụdị Whisper Voice-to-text.

Anyị ga-enye iwu nke a ga-edegharị, gbanwee ka ọ bụrụ ederede, jiri Whisper wee nyochaa iji mee usoro kwesịrị ekwesị, nke nwere ike ịbụ site na ịme mmemme ruo inye voltaji na ntụtụ RaspberryPi.

M ga-eji Raspberry Pi 2 ochie, micro USB na m ga-eji ụdị olu-na-ederede nke OpenAI wepụtara na nso nso a, Na-agba ọsọ. Na njedebe nke isiokwu ị nwere ike ịhụ ntakwu ntakwu ntakwu.

emebere ha niile Python.

Idenye aha na ọwa Youtube anyị

M na-ahapụrụ gị ihe ngosi nke otú o si arụ ọrụ na vidiyo a, na-achịkwa PC site na olu.

Mgbakọ

Iji jiri ya na PC, anyị ga-achọ naanị igwe okwu.

Ọ bụrụ na ị ga-ebunye ya na RaspberryPi, ị ga-achọ igwe okwu USB, n'ihi na jakị ọ nwere bụ naanị maka mmepụta.

Mkpa:

Dị ka ebumnuche izugbe nke ngwá ọrụ bụ njirimara olu. Achọpụtara m na ọ bara ezigbo uru itinye ya n'ọrụ nke ngwaọrụ ndị ọzọ.

Micro USB
Raspberry PI nwere sistemụ arụmọrụ (ihe atụ Raspbian pro)
Eletrọnịkị (LED, wires, resistor 480 ohm na bred)

Anyị na-ejikọta LED na pin 17, nke bụ nke anyị ga-arụ ọrụ ma gbanyụọ maka ahụmahụ a.

mmepe koodu

A na-ekewa ya n'ime akụkụ atọ, nke mbụ, ihe ndekọ ọdịyo nke m nwetara koodu na ya geeksforgeeks, n'ihi na amaghị m ụlọ ahịa akwụkwọ ndị ahụ. Nke abụọ, ntụgharị nke ọdịyo gaa na ederede na Whisper na nke atọ, ọgwụgwọ nke ederede na nzaghachi na RaspberryPi

N'ihe atụ ule, naanị m ga-emekọrịta ihe na Led, na-eme ka ọ na-enwu ma ọ bụ na-egbuke egbuke, mana anyị nwere ike ịmepụta edemede iji dozie ya na mkpa anyị.

Ama m na nke a bụ Raspberry Pi 2 na ọ ga-adị nwayọ karịa Raspberry Pi 4, mana maka ịnwale ọ dị mma.

Tupu ị nweta ya na-arụ ọrụ, ị ga-achọ ịwụnye ihe ndị a

#Instalar whisper
pip install git+https://github.com/openai/whisper.git
sudo apt update &amp;&amp; sudo apt install ffmpeg

#para que funcione la grabación de audio
python3 -m pip install sounddevice --user
pip install git+https://github.com/WarrenWeckesser/wavio.git

#si vas a instalarlo en la raspberry
#dar permisos para usar la GPIO
sudo apt install python3-gpiozero
sudo usermode -aG gpio <username>

koodu niile

#!/usr/bin/env python3
import whisper
import time
from gpiozero import LED
import sounddevice as sd
from scipy.io.wavfile import write
import wavio as wv

        
def main ():
    inicio = time.time()
    record_audio ()

    model = whisper.load_model("tiny")
    result = model.transcribe("audio1.wav")
    words = result["text"].split()

    for word in words:
        word = word.replace(',', '').replace('.', '').lower()
        if word == 'enciende' or 'encender':
            encender()
            break
        if word == 'parpadea' or 'parpadear':
            parpadear()
            break      
    fin = time.time()
    print(fin-inicio)

def encender ():
    LED(17).on()

def parpadear ():
    light = LED(17)
    while True:
        light.on()
        sleep(1)
        light.off()
        sleep(1)

def record_audio ():
    # Sampling frequency
    freq = 44100
    # Recording duration
    duration = 5
    # Start recorder with the given values
    # of duration and sample frequency
    recording = sd.rec(int(duration * freq),
                    samplerate=freq, channels=2)
    # Record audio for the given number of seconds
    sd.wait()
    # This will convert the NumPy array to an audio
    # file with the given sampling frequency
    write("audio0.wav", freq, recording)
    # Convert the NumPy array to audio file
    wv.write("audio1.wav", recording, freq, sampwidth=2)
        
main ()


#dar permisos para usar la GPIO
#sudo apt install python3-gpiozero
#sudo usermode -aG gpio <username>

#Instalar whisper
#pip install git+https://github.com/openai/whisper.git
#sudo apt update &amp;&amp; sudo apt install ffmpeg

Enwebeghị m ike ịnwale ya n'ihi na enweghị m microSD maka RaspberryPi, ma ọ bụ igwe okwu USB jikọọ, mana ozugbo m nwara ya, m na-edozi njehie ụfọdụ na ọ dị mfe ịbanye.

Nzọụkwụ site nzọụkwụ nkọwa nke koodu

#!/usr/bin/env python3

Shebang ga-agwa ngwaọrụ ahụ asụsụ anyị mebere na ya na onye ntụgharị okwu anyị ga-eji. Ọ bụ ezie na ọ dị ka ihe na-adịghị mkpa, ọ bụghị itinye ya na-akpata mmejọ ọtụtụ oge.

ọba akwụkwọ ndị ebubata

import whisper
import time
from gpiozero import LED
import sounddevice as sd
from scipy.io.wavfile import write
import wavio as wv

Tụgharịa na-arụ ọrụ na ihe nlereanya

oge, n'ihi na m na-eji ya chịkwaa oge ọ na-ewe iji mebie script, gpiozero na-arụ ọrụ na GPIO ntụtụ nke Raspberry na sounddevice, scipy na wavio ịdekọ ọdịyo.

Ọrụ

Emeela m ọrụ anọ:

isi ()
ìhè ()
kpuo ìsì ()
record_audio()

gbanye () naanị na-enye voltaji na pin 17 nke rasberi ebe anyị jikọtara na nke a LED iji nwalee.

def encender ():
    LED(17).on()

blink () dị ka ịgbanwuo () mana ọ na-eme ka ndị na-edu ndú na-egbuke egbuke site na ịgbanwuo ya n'ime akaghị.

def parpadear ():
    light = LED(17)
    while True:
        light.on()
        sleep(1)
        light.off()
        sleep(1)

Site na record_audio() anyị na-edekọ faịlụ ọdịyo

def record_audio ():
    # Sampling frequency
    freq = 44100
    # Recording duration
    duration = 5
    # Start recorder with the given values
    # of duration and sample frequency
    recording = sd.rec(int(duration * freq),
                    samplerate=freq, channels=2)
    # Record audio for the given number of seconds
    sd.wait()
    # This will convert the NumPy array to an audio
    # file with the given sampling frequency
    write("audio0.wav", freq, recording)
    # Convert the NumPy array to audio file
    wv.write("audio1.wav", recording, freq, sampwidth=2)

Isi bụ isi ọrụ, rịba ama na naanị ihe anyị nwere na mpụga nke ọrụ bụ oku na-aga na isi () na njedebe nke edemede ahụ. N'ụzọ dị otú a na mmalite, ọ ga-ebubata ọba akwụkwọ wee kpọọ ọrụ ahụ.

def main ():
    inicio = time.time()
    record_audio ()

    model = whisper.load_model("tiny")
    result = model.transcribe("audio1.wav")
    words = result["text"].split()

    for word in words:
        word = word.replace(',', '').replace('.', '').lower()
        if word == 'enciende' or 'encender':
            encender()
            break
        if word == 'parpadea' or 'parpadear':
            parpadear()
            break      
    fin = time.time()
    print(fin-inicio)

Anyị na-echekwa oge anyị na-amalite ịrụ ọrụ ahụ wee kpọọ ọrụ ndekọ ndekọ nke ga-edekọ ntụziaka anyị na faịlụ .wav, .mp3, wdg nke anyị ga-emesị gbanwee gaa na ederede.

    inicio = time.time()
    record_audio ()

Ozugbo anyị nwetara ọdịyo ahụ, a ga-akpọ ntanye ntanye anyị gwa ya ihe nlereanya anyị chọrọ iji, enwere 5 dị, anyị ga-eji obere obere, n'agbanyeghị na ọ bụ nke kachasị njọ n'ihi na ọ kachasị ọsọ na ọdịyo ga-adị mfe. naanị okwu 3 ma ọ bụ 4.

     model = whisper.load_model("tiny")
    result = model.transcribe("audio1.wav")

Site na nke a, anyị na-atụgharị ụda ka ọ bụrụ ederede ma chekwaa ya na mgbanwe. Ka anyị gbanwee ya ntakịrị.

Anyị na-atụgharị nsonaazụ ka ọ bụrụ ndepụta na okwu ọ bụla nke ọdịyo ahụ

     words = result["text"].split()

Na ihe niile dị njikere imekọrịta na ngwaọrụ anyị. Ugbu a, anyị ga-emepụta naanị ọnọdụ anyị chọrọ.

Ọ bụrụ na ọdịyo nwere okwu X, mee Y. Dị ka anyị nwere okwu na ndepụta, ọ dị mfe itinye ọnọdụ

         for word in words:
        word = word.replace(',', '').replace('.', '').lower()
        if word == 'enciende' or 'encender':
            encender()
            break
        if word == 'parpadea' or 'parpadear':
            parpadear()
            break

Ahịrị

         
        word = word.replace(',', '').replace('.', '').lower()

M na-eji ya ịtụgharị mkpụrụokwu ndị dị na ọdịyo ka ọ bụrụ obere ma wepụ akararị rịkọm na oge. Na n'ụzọ dị otú a zere njehie na ntụnyere

Na nke ọ bụla ma ọ bụrụ na ọnọdụ nke inwe nke ọ bụla n'ime okwu ndị anyị họọrọ emezu, ọ na-akpọ ọrụ nke ga-eme ihe anyị chọrọ;

Nke a bụ ebe anyị na-agwa ya ka ọ rụọ ọrụ PIN nke ga-enwu ọkụ ma ọ bụ mee ka ọ nwuo. Gbaa ụfọdụ koodu, ma ọ bụ mechie kọmputa.

Ihe a niile bụ echiche bụ isi. Site ebe a ị nwere ike ịmepụta ọrụ ahụ ma melite ya dịka ịchọrọ. Onye ọ bụla nwere ike ịchọta ụzọ dị iche iche maka ya.

Ihe anyị nwere ike iji montage a mee

Ndị a bụ echiche na-abịakwute m iji nweta uru nke montage a. Ozugbo ọkpụkpụ ahụ na-eji ngwá agha, anyị nwere ike iji ya mee ihe ọ bụla na-abata n'uche site na olu, anyị nwere ike ịgbalite relay nke na-amalite moto ma ọ bụ anyị nwere ike ịmalite edemede nke na-eme edemede, email ma ọ bụ ihe ọ bụla.

Gịnị bụ ntakwu

Whisper bụ ụdị njirimara vol, na-arụ ọrụ n'asụsụ dị iche iche yana ọnụ ọgụgụ asụsụ dị ukwuu ma na-enye ohere ntụgharị n'ime Bekee. Ọ bụ ihe anyị maara dị ka ngwá ọrụ ederede na olu, nke ndị otu OpenAI wepụtara, ndị okike nke Dall-e.