Preserving Voices from the Noise: How LifeOnRecord Uses LALAL.AI API at Scale

LifeOnRecord processes over 100,000 minutes of audio a year, helping people to preserve their dearest moments on weddings, anniversaries, birthdays, retirements, and other vital milestones of their lives, with every recording coming from a real event. Each with its own issues that affect how clear the voice sounds: The DJ might be louder than the guest leaving the message or someone might be yelling into the phone just to be heard over the crowd. And yet, every one of those recordings carries something irreplaceable.

For this case study, we sat down with Alaa El Ghatit, founder of LifeOnRecord, to talk about how API-powered automation helped a family business handle more recordings without losing the personal care the product depends on, and how LALAL.AI was brought in to make the memorable voices easier to hear.

LifeOnRecord: Family Business That Helps Preserve the Happiest Moments 

LifeOnRecord collects voice messages from family and friends for weddings, anniversaries, birthdays, retirements, memorials, Bar Mitzvahs, celebrations of life, family reunions, graduations, and other milestone events. We’re also used at corporate conventions, conferences, brand activations, and experiential events where organizations want to collect authentic voices and reactions from participants.

The service has also been used in high-profile settings involving entertainers, athletes, public figures, and even the pope, but the emotional core is the same whether the event is large and public or small and private: someone wants to preserve voices from a meaningful moment.

The work begins before the event. Customers create an account on LifeOnRecord, get access to an online portal, a dial-in number, and an invitation number. The dial-in side is active right away, which means friends and family can start leaving messages before the event. We support dial-in numbers in more than 30 countries; people can also leave messages through WhatsApp, email recordings, or upload files to us. Those files are automatically routed to the correct account and added to their portal.

Over time, that expanded into modified vintage and retro phones for in-person events: Guests pick up the handset, hear a greeting, leave a message, and hang up.

We’ve selected retro phones as they look beautiful at an event, fit naturally into wedding decor, and they instantly signal to guests that this is something different from signing a book or posing for a camera. Written notes are wonderful, but a voice captures something different: laughter, timing, hesitation, emotion, accent, personality, and the way someone tells a story. Those details are hard to preserve in any other format.

After an event, the physical phone comes back to us. We then route those recordings to the right account and add them to the customer's online portal, so everything becomes part of one collection. 

Customers can then stream the recordings from any smartphone, tablet, or computer, download the messages as individual MP3s or as one long stitched-together MP3, and, if customers wish, preserve the recordings on a keepsake speaker or vinyl record.

Our clients also receive written transcriptions of all the recordings, which matters because the text makes the collection easier to scan, search, and revisit. We send the audio for automated transcription after it has gone through human editing and LALAL.AI processing, so the transcription engine is working from cleaner audio.

That is why the quality of the audio matters: these are not just audio files but people's voices at meaningful moments.

“For me, the business is about making it easy to preserve voices that might otherwise disappear after the moment passes.”

We’ve saved every recording that has ever been left with LifeOnRecord since the business began about 20 years ago, which has allowed us to help customers recover recordings they thought they had lost.

It’s Technically Challenging to Get Clear Audio Recordings at Weddings 

A wedding reception is not a studio: There may be loud music, cheering, clinking glasses, guests talking over one another, kids playing with the phone, people moving around, and different guests speaking at very different volumes. We’re not trying to make a wedding sound like a studio; we’re trying to make the voices easier to hear while preserving the feeling that the message came from a real celebration. The goal is clarity without making the recording sound artificial or overprocessed.

The venue itself matters. A ballroom, barn, tent, outdoor space, hotel lobby, or reception hall can all create different audio problems. Some spaces are echoey. Some have constant background music. Outdoor events can add wind, traffic, or other ambient noise. During the same event, the sound can change from quiet dinner conversation to loud dancing within an hour.

There have been many times when guests leave messages saying things like, 'I do not know if you can hear this,' because the music is so loud around them. In those situations, the emotional content may be wonderful, but the raw recording can be difficult to listen to. 

“That is where strong audio processing becomes valuable, and it's amazing how background noise can just fall off after correct processing with LALAL.AI.”

“We have an 'enhance early in the process' philosophy.”

That means the audio is improved before it's used for streaming, downloads, transcription, or keepsakes. We do not wait until someone orders a speaker or vinyl record to enhance the audio; the recordings customers hear in the portal and can download are already the cleaned, reviewed versions.

Before we had a more automated workflow, the process was much more manual. We'd open recordings in desktop tools, inspect waveforms, identify obvious problems, trim silence, adjust volume, reduce noise where possible, and export the finished files.

That can work when there are only a few recordings, but audio guestbooks create a lot of short files. Across the business, we process a large amount of audio annually, roughly in the range of 100,000 minutes a year. In many weeks, we are processing more than 1,000 individual recordings and more than 1,000 minutes of audio. Such volumes mean we need automation, but the customer experience still depends on care, review, and judgment.

“With LALAL.AI, the value is that the processing can be used programmatically rather than as an occasional manual tool."

We can apply it as part of a repeatable workflow, especially for the physical phone recordings where the audio environment is most unpredictable.

We still review edge cases manually, but for reducing noise, music, and distractions around a recoverable voice, API-based processing gives us a dependable first pass. Automation is one of the reasons we can support more events without the workflow breaking down but I do not think of scaling as becoming a factory. Scaling for us means being able to serve more customers while preserving the care that makes the product meaningful. 

“LALAL.AI is especially valuable for physical phone recordings from live events. Those recordings are the least controlled.”

Before LALAL.AI, we had used another API-based processing approach. It was better than manual processing, but it required a lot of parameter decisions: loudness targets, dynamics, noise reduction, filtering, speech isolation, click reduction, plosive reduction, sibilance reduction, and other settings.

The problem was not that those controls were bad but that our audio environment is highly variable. 

When I tested LALAL.AI head-to-head, I was pleased by how few parameters were needed and how strong the results were. It felt more dynamic and flexible for the variety of recordings we handle.

“What stood out was that LALAL.AI seemed to produce better overall quality and performance without requiring us to constantly guess which combination of settings would be right for each file.”

LALAL.AI is especially valuable for physical phone recordings from live events. Those recordings are the least controlled and the most likely to include DJ music, crowd noise, room echo, handling noise, and people speaking at inconsistent volumes.

One very clear example was an event where someone pulled the handset cord and damaged it. After that, a very loud buzzing was introduced into the recordings, almost drowning out the voices. LALAL.AI removed the buzzing so completely that the voices became clear and you could no longer tell anything had gone wrong.

Another common scenario is a phone placed right next to the dance floor. We’ve had recordings where the DJ music is louder than the person leaving the message. Sometimes guests are yelling into the phone and saying, 'I don't know if you can hear this.' After LALAL.AI, the voice can become crystal clear, with no background music distracting from the message. 

That is the ideal use case for us: the voice is there, but the event environment is getting in the way. LALAL.AI helps the message come through and reduce the distractions around the voice so the recordings are easier and more enjoyable to hear.

We use different LALAL.AI settings depending on the source. For the physical phones, Andromeda with de-echo turned off has been a strong fit. For other sources, such as dial-in recordings or greetings, the processing path can be tuned differently because the recordings usually start out cleaner and have different characteristics.

The key is that LALAL.AI is not just a separate tool we open once in a while. It's part of an integrated workflow that helps us process real-world audio at scale, enhance early for portal delivery and transcription, preserve the original recordings, and still review edge cases manually. 

Unlike other APIs, LALAL.AI feels more dynamic and less dependent on us choosing a long list of fixed parameters. With some audio processing tools, there are many settings to adjust, and it can be difficult to know whether a particular combination is best for a recording.

“What comes out the other end of LALAL.AI seems magical: voices are clear with no background noise interference.”

With LALAL.AI, the smaller number of parameters has actually been a strength. It has produced strong consistent results without making us micromanage every file. The voice-focused cleanup has been particularly useful when music or noise competes with the speaker.

I would not say that any API solves every audio problem. The best workflow is still a combination of automated processing, testing, routing, and human review. But LALAL.AI gets closer than anything else and has become a very useful part of that toolkit.

The Best Technology Disappears & Lets the Message Come Through

I get the most satisfaction when our product is used by friends and family who wish to support someone with a terminal illness, or for someone celebrating an older milestone birthday, such as a 70th or 80th. In those situations, people may be scattered across the country or around the world, and the act of leaving a message can reconnect them in a way that might not have happened otherwise.

That is where I want the business to go. Automation helps us handle the logistics, routing, audio enhancement, transcription, and delivery, so we can spend more energy on the human side of the experience. 

The goal is not simply to process more files but to help people preserve more meaningful voices.

For businesses like ours, LALAL.AI is valuable because it helps turn messy real-world event recordings into something customers can enjoy and keep. These are not just audio files. They are people's voices at meaningful moments. 

The best technology disappears and lets the message come through.

Follow LALAL.AI on Instagram, Facebook, Twitter, TikTok, Reddit, LinkedIn, and YouTube to keep up with all our updates and special offers.