Using AI to create a multi-lingual answering machine announcement for your FRITZ!Box

Recording a message, even in a single language, can be difficult. Many of us cringe at the sound of our own voice played back to us. Furthermore, it takes practice to record a message without interruptions while maintaining a steady pace and clear enunciation.

The complexity multiplies when you’re looking to create a multi-lingual answering machine announcement. Fortunately, AI is making this a lot easier than it used to be.

Write the text for the announcement

An answering machine on a FRITZ!box typically allows you to record 3 types of messages:

  1. A greeting for the “Record messages” operating mode.
  2. A “goodbye message” that plays when the caller reaches the maximum recording time.
  3. An announcement for the “Greeting only” operating mode, where callers cannot leave a message.

Start by writing the message(s) in your native language. If you’re not sure what to write or if you want to be particularly funny or obnoxious, you can always ask ChatGPT for help. Here’s a suggestion:

Hey, you've reached my voicemail. I'm currently on vacation and as unreachable as your dream of becoming a millionaire. I'll be back soon, so try again later. Cheers, you ambitious bugger!

Translate the announcement

If you need help with the other language versions, use DeepL translator, Google Translate or similar websites to translate the text:

Use ElevenLabs Voice AI for voice synthesis

If you want your announcement to sound like a robot, click on the speaker icon in Google Translate and record the output (actually it does not sound that bad anymore, so this may have become a viable option). However, for the highest-quality, human-like voice synthesis and a choice of different voices, there’s currently no better option that Prime Voice AI by ElevenLabs.

In my experience, some voices sound better for certain languages than others. However, Prime Voice AI is evolving rapidly, so I won’t go into details as they will be quickly outdated.

Download recordings of your translated texts in all the desired languages (these will be MP3 files):

ElevenLabs Prime Voice AI


If you’re creating an announcement for personal rather than business use, consider making it clear to the caller that they’ve reached the right person, even if the recording doesn’t sound like you. On the other hand, if you like to confuse people, this could be a great opportunity.

ElevenLab’s free plan requires “attribution to”, though I’m not sure how this would work with a pure audio recording. Perhaps you could incorporate it into your message, something like: “This is Adam from I’m sorry, but <your name> cannot pick up the phone right now…”

Processing the audio files in Audacity

Open Audacity and drag-and-drop the MP3 files you downloaded from ElevenLabs. They will appear as separate tracks:

Audacity window with 3 tracks

Normalize the clips (when using different voices)

If you notice a significant volume discrepancy between the tracks, normalize them first to even out the volume levels before you combine them. Normalizing will ensure that each individual track reaches its maximum volume level without distortion.

  1. Press Ctrl+A to select all the tracks or go to “Select” > “All”.
  2. Go to “Effect” > “Volume and Compression”> “Normalize”.
  3. I’ve found that the normalizing to -3.0 dB works well for the FRITZ!box. Click “Apply”.

Move the clips to one track

You may want to “Zoom out” (Ctrl + 3) to make this easier. Then grab the clip you want to move at it’s light-colored area at the top and drag it onto the top track (behing the clip that is already on the track, leaving some room for a pause, if required). Repeat this process for the remaining clip(s).

If you want, you can now delete the empty tracks by clicking on the ‘x’ at the top left of each track.

Audacity will 3 clips on the same track

Export in the best format for the FRITZ!box answering machine

Your FRITZ!Box will likely prefer recordings in WAV format, 8000 Hz, 16-bit, and mono. However, it’s best to confirm this in your device’s manual or the FRITZ!box knowledge base.

The MP3 file downloaded from Elevenlabs should already be mono. If not, go to “Tracks” > “Mix” > “Mix Stereo Down to Mono”.

When exporting the file as “WAV (Microsoft) signed 16-bit PCM”, Audacity automatically converts the bit depth to 16-bit, regardless of your project’s current setting. So, there’s no need to change this.

However, we need to change the sample rate to 8000 Hz. One way to do so is by resampling the track:

  1. Go to the “Tracks” menu at the top of the screen.
  2. In the dropdown menu, select “Resample…”.
  3. In the “Resample” dialog box that pops up, enter “8000” in the “New sample rate (Hz)” field.
  4. Click “OK” to apply the changes.

Now, you can listen to the announcement at the much lower sample rate to get an idea of how it’ll sound to your callers.

Finally, go to “File” > “Export” > “Export as WAV”. In the dialog box that appears, enter your desired file name, make sure “WAV (Microsoft) signed 16-bit PCM” is selected as the type, and click “Save”.

If you also want to preserve your Audacity project file, press Ctrl + S or go to “File” > “Save Project”.

Import the audio file into your Fritz!box

The process of setting up an answering machine on your FRITZ!Box may vary based on your specific model and firmware. However, within the answering machine’s settings, you should find an option to use your own announcement instead of using the standard greeting. Here, choose the .WAV file that you exported from Audacity:

For further information, please search for “Configuring Your Own Greetings” in the FRITZ!box knowledge base.

Solution: Cannot access QNAP NAS from Windows

The problem: You cannot access the SAMBA (SMB) shares on your QNAP NAS from the Windows File Explorer and after you tried, access to the web interface also stopped working. The NAS is still accessible from other PCs on the same network.

The reason: Windows will first try to connect to your NAS using your Windows login credentials. If this fails too many times and Network Access Protection on your NAS is enabled, your PC’s IP address will be banned:

QNAP Network Access Protection

You can see the failed login attempts in the System Connection Logs (if enabled for SAMBA):

QNAP System Connection Logs


There are many threads about this issue on the QNAP forum proposing various solutions, such as:

  • Creating a user account on the NAS with the same user name and password as on Windows.
  • Disabling Network Access Protection on the NAS.
  • Making all folders public.
  • Disabling all guest shares on the NAS to force Windows to show the network credentials dialog.

I tried a different solution which I found here. I left everything on the NAS unchanged. In the local group policy settings of my Windows PC, I enabled insecure guest logons for the SMB client. This allows the PC to connect to SAMBA shares which are not protected by a password (such as the Public folder on a QNAP NAS).

Note: The Group Policy Editor (gpedit.msc) is not available in all versions of Windows 10 and Windows 11. It is primarily included in the Professional, Enterprise, and Education editions of these operating systems. However, you can use a tool called ‘Policy Plus’ to add it to the home edition.

To start the Local Group Policy Editor, press your Windows key and start typing “group policy” (then select “Edit group policy”):

Alternatively, you can also press the Windows key and R, then type “gpedit.msc” and click on OK:

In the Local Group Policy Editor, the setting can be found under Local Computer Policy -> Computer Configuration -> Administrative Templates ->Network -> Lanman Workstation -> Enable insecure guest logons:

Windows Registry editor

Note: The description says that “if you do not configure this policy setting, the SMB client will allow insecure guest logons.” This was not true for me on Windows 11 21H2, the setting had to be explicitly enabled. On other PCs running Windows 10, the described behavior was correct and the entire problem never appeared.

Now (and after the IP ban period has expired, of course) it was possible to access the NAS through the file explorer. Clicking on a protected folder brought up the Enter network credentials dialog:

Enter network credentials

Why did this work? Apparently, Windows only tried to use the current user credentials once when connecting to the NAS and then used guest access. As this resulted only in a single failed access attempt, the PC’s IP address was not banned.

For further support, please visit the QNAP forum

Update February 15, 2023: A previous version of this article mentioned making changes to the registry, however the screenshot showed the Local Group Policy Editor (gpedit.msc) and not the registry editor (regedit.exe). Many thanks to Daniel for the comment.

Update March 28, 2023: Added a clarification that the Group Policy Editor (gpedit.msc) is only available on certain editions of Windows and changed the title accordingly.

Easy CPU cooler replacement for Acer PO3-620 i7 10700

Buying Acer’s Predator Orion gaming PC is one way to get an RTX 3070 graphics card without paying absurd prices. However, it comes with a standard CPU cooler that is loud and not very effective.

I replaced it with a Xilence XC026| I4026 ( / affiliate links), which led to a significant CPU temperature drop under load. While this doesn’t turn the PC into a quiet machine, the cooler can be installed without replacing the backplate.

All you have to do is modify the cooler itself by moving the two brackets that hold the large screws as shown below.

Original configuration

CPU cooler in original configuration
Bracket at the bottom of the cooler

After modification

Modified CPU cooler
Bracket after being moved (ignore the incorrect orientation of the bracket the back, I wasn’t paying attention and also might be stupid)

On one side you can remove the fan to fasten the small bracket screws. On the other side there is very little room (this is the only challenge with this procedure):

Very little room to access screws on one side
Still easier than removing the motherboard to replace the backplate

Follow the instructions that came with the cooler for the rest of the (simple) installation process.

CPU temperature at full load

Temperature after 30 minutes of Folding@home, ambient temperature 24°C:

I found the recommendation for this cooler on the German Acer forum.

Driver for LogiLink AU0033 USB 2.0 to 8x Serial Adapter

You can download it here.

How I found it

I went to the product page on the LogiLink website, clicked on “Downloads”, then on the link to the driver which sent me to a 404 page on

I then searched for the product directly on the website, found it here, clicked on “Downloads”, then on “Treiber AU0033” (Treiber is German for driver), which led to a PDF document that in turn contained the link to the actual driver ZIP file (which is hosted on the LogiLink website where I started).

Note if you’re reinstalling the driver

You can find an uninstaller under “Logilink AU0033 8x seriell\CD\Driver\Windows\64X\9710_7840_QUADPORT_MSUninst.exe”, running this before the installer fixed a problem on my system where only half of the ports would appear.

Connecting a scale or balance to a Raspberry Pi: project ideas

On, I showed how easy it can be to send the weight from a scale to a Python script running on a Raspberry Pi (if you choose the right scale). As I had never used Python before, all the script did was print the weight data. That doesn’t mean you have to stop there. Here a some ideas and links to get you going.

  • Use the Raspberry Pi as a serial device server (from RS-232 to TCP/IP over Ethernet or WiFi). Code examples can be found in the pySerial documentation.
  • Build a protocol converter. Most scales use proprietary protocols, but you can convert the data to MQTT or other protocols as shown here. Do not say that you have developed a protocol converter, call it an IoT gateway to get more attention.
  • Use speech synthesis to output the weight as spoken words for visually impaired users.
  • Use speech recognition to send commands to the scale (e.g. the tare command).
  • Install a Pi Camera Module and take a picture of the user every time the scale is overloaded.
  • Turn the scale into a checkweigher by comparing the weight with preset values and making under/accept/over LEDs light up.
  • Transform the weight into a proportional analog signal (voltage) because… I have no idea! Why do people keep asking for weighing scales with an analog output? Seriously, if you know the answer, please tell me.
  • Trigger an alarm or something worse when an object is removed from the scale: