Tag Archives: Google

Transcribe audio with Google Cloud speech-to-text api

I had a few audio files of an interview done with a late relative that I wanted to have Google transcribe for me. I wanted to supply an audio file and have it spit out the results. There are many ways to do this but I went with using the Google Cloud Platfrom speech-to-text API.

First I signed up for a GCP free trial via https://cloud.google.com/speech-to-text/ For my usage, it will remain free as 0-60 minutes of transcription per month is not charged: https://cloud.google.com/speech-to-text/pricing

Next, I needed to create GCP storage bucket as audio more than 10 minutes long cannot reliably be transcribed via the “uploading local file” option. I did this following the documentation at https://cloud.google.com/storage/docs/creating-buckets which walks you through going to their storage browser and creating a new bucket. From that screen I uploaded my audio files (FLAC in my case.)

Then I needed to create API credentials to use. I did this by going speech API console’s credentials tab and creating a service account, then saving the key to my working directory on my local computer.

Also on said computer I installed google-cloud-sdk (on Arch Linux in my case, it was as simple as yay -S google-cloud-sdk)

With service account json file downloaded & google-cloud-sdk installed I exported the GCP service account credentials into my BASH environment like so

export GOOGLE_APPLICATION_CREDENTIALS=NAME_OF_SERVICE_ACCOUNT_KEYFILE_DOWNLADED_EARLIER.json 

I created .json files following the format outlined in command line usage outlined in the quickstart documentation. I tweaked to add a line “model”: “video” to get the API to use the premium Video recognition set (as it was more accurate for this type of recording.) This is what my JSON file looked like:

{
  "config": {
      "encoding":"FLAC",
      "sampleRateHertz": 16000,
      "languageCode": "en-US",
      "enableWordTimeOffsets": false,
      "model": "video"

  },
  "audio": {
      "uri":"gs://googlestorarge-bucket-name/family-memories.flac"
  }
}

I then used CURL to send the transcription request to Google. This was my command:

curl -s -H "Content-Type: application/json" -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) https://speech.googleapis.com/v1/speech:longrunningrecognize -d @JSON_FILE_CREATED_ABOVE.json

If all goes well you will get something like this in response:

{
  "name": "4663803355627080910"
}

You can check the status of the transcription, which usually takes half the length of the audio file to do, by running this command:

curl -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) -H "Content-Type: application/json; charset=utf-8" "https://speech.googleapis.com/v1/operations/ID_NUMBER_ACQUIRED_ABOVE"

You will either get a percent progress, or if it’s done, the output of the transcription.

Success! It took some time to figure out but was still much better than manually transcribing the audio by hand.

Manually update firmware on Nexus devices

The release of Android 6.0.1 had me excited because it enables LTE band 12 for my phone, the Nexus 5X, which currently uses T-mobile. Band 12 is in the 700mhz range which should greatly increase speed and coverage. I’m too impatient to wait for the OTA!

This tutorial will walk you through how to manually backup, unlock, flash, re-lock, and restore a Google Nexus 5X, but the procedure is pretty much the same for any Nexus device.

First, obtain two necessary Android development tools: adb and fastboot. Do not use your distribution’s versions of these tools – they are likely out of date. Instead, download the Android SDK directly from Google by going here and scrolling to the bottom of the page. Java is required for the sdk to install – install it if you haven’t already. Thanks to this site for explaining how to only obtain platform-tools.

sudo apt-get install openjdk-7-jre
tar zxvf android-sdk_r24.4.1-linux.tgz
android-sdk-linux/tools/android update sdk --no-ui --filter 1,platform-tools

Once you have platform-tools you need to add them to your PATH to make scripts run adb successfully (thanks to this site for the information). When you run the command below make sure to update /path/to/… to the folder where you extracted the android-sdk.

echo "export PATH=$PATH{}:/path/to/android-sdk/tools:/path/to/android/platform-tools" >> ~/.bashrc
source ~/.bashrc

Next, obtain the latest firmware for your device from the google developers site. Extract it somewhere you will remember for later.

Now, enable USB debugging on your phone if you haven’t already (thanks to this site for the info.) To do this, go to Settings / About phone, scroll to the bottom and press on the build number 7 times. Press back and go to developer options, and enable USB debugging and enable OEM unlocking. Then plug in your phone to the computer with adb installed and run this command (thanks to xda for the information)

adb backup -apk -shared -all -f <backup_filename>

Confirm the on-screen prompt on your phone. Make sure you specify a password for encryption. The above command should backup everything, but in my case it did not backup files in the flash partition (downloads, pictures, etc). Make sure you manually copy any important files from your phone before you proceed.

My first attempt at the above command didn’t work for me. I received the error message:

adb: unable to connect for backup

When I ran adb devices it showed this:

List of devices attached 
00c739918fbf4e2a offline

It turns out I had an old version of adb installed. Make sure you download Google’s official SDK instead of relying on your distribution’s version.

Once the backup is complete, you then need to reboot your phone into fastboot mode:

adb reboot bootloader

Wait for the reboot, then run the following command. Warning: this command will wipe your device. Make sure you have a reliable backup and confirm the message on your phone screen.

fastboot oem unlock

Now, navigate to the directory where you extracted your latest firmware and execute the flash-all script:

cd bullhead-mmb29k
./flash-all.sh

After some time your phone will reboot into your shiny new updated OS. Skip everything setup-wise. Re-enable developer mode and android debugging, and then re-lock your bootloader:

adb reboot bootloader
fastboot oem lock

Lastly, we need to restore everything from the backup we made and re-lock the bootloader for security. Once again skip everything setup-wise on the phone, re-enable developer mode and android debugging, and then restore your stuff:

adb restore <filename>

Don’t forget to manually copy back any flash files you manually backed up earlier.

The very last (optional) step is to go into developer options settings and disable OEM unlocking.

Success!