LALAL.AI

🟡 Andromeda is now available on LALAL.AI desktop and mobile applications! (Stems: Vocal & Instrumental and Voice & Noise) And yes, it's accessible to everyone by default and included in your current plan. Regardless of the device you're running LALAL.AI on, you can now enjoy: — Up to 40% faster processing, enabling more efficient workflows for creators and audio teams. — 10% improvement in SDR, resulting in cleaner, more accurate separations. — The frequency range remains consistent at up to 22 kHz with stereo processing, providing high-fidelity separation the industry expects. Andromeda handles challenging audio scenarios and complex instrument mixes with greater precision, reducing artifacts and increasing consistency — Thanks to advanced DSP (Digital Signal Processing) techniques, Andromeda delivers consistent separation across quiet or loud tracks, letting creators and audio engineers work across different types of audio sources without worrying about volume inconsistencies. — Reduced need for manual DAW cleanup, thanks to more precise layer separation and improved handling of complex mixes. — No more trade-offs between detail and bleed control. Andromeda eliminates the former Clear Cut and Deep Extraction modes, giving you clean, detailed stems in a single pass. Just make sure your app's version is 2.11 🤍 And share your feedback on how the new algo works for you!

Andromeda Removes a Long-Standing Trade-Off in Stem Separation Our newest, fastest, and clearest algorithm to date, Andromeda, removes the need to choose between Clear Cut and Deep Extraction, delivering detailed stems with minimal bleed in a single pass, fewer processing runs, less DAW cleanup, faster deliveries for teams and studios. Read more in the carousel below & try Andromeda on LALAL.AI to test its stem separation quality.

🎧 Introducing Andromeda: LALAL.AI’s Most Advanced Separation Model to Date & New Benchmark in Stem Separation We’re pleased to announce the release of Andromeda, the latest evolution of our audio source separation technology that interprets audio tracks with near-human precision. Built on six generations of research and extensive DSP (Digital Signal Processing) work, Andromeda takes a more nuanced approach, analyzing audio in terms of time, frequency, and tone, rather than simply processing sound as a single waveform. Key improvements that matter: • Up to 40% faster processing, enabling more efficient workflows for creators and audio teams. • 10% improvement in SDR, resulting in cleaner, more accurate separations. • The frequency range remains consistent at up to 22 kHz with stereo processing, providing high-fidelity separation the industry expects. • Thanks to advanced DSP techniques, Andromeda delivers consistent separation across quiet or loud tracks. • Reduced need for manual DAW cleanup, thanks to more precise layer separation and improved handling of complex mixes. • No more trade-offs between detail and bleed control. Andromeda eliminates the former Clear Cut and Deep Extraction modes, giving users clean, detailed stems in a single pass. Andromeda is now available across key LALAL.AI tools, including Stem Splitter (Vocal & Instrumental and Voice & Noise stems), Lead & Back Vocal Splitter, Echo & Reverb Remover, and Voice Cleaner. The new algorithm is also available via API for Stem Splitter & Voice Cleaner clients. 💜 Black Friday Special To mark the release, we’re offering a limited-time Black Friday Sale valid until December 5. Don't miss the chance to celebrate the release with us!

💡Do you know that not long ago, YouTube introduced automatic dubbing, a feature designed to help creators connect with audiences across different countries? If you’ve come across videos dubbed this way, you’ve probably noticed how good the translation is. At the same time, it’s hard not to notice how unnatural or even robotic these dubbed voices often sound. It’s especially noticeable when you’re watching a dubbed video from a creator whose real voice you’ve known for years before the auto-dubbing feature was introduced. The difference can feel jarring. The good news is that creators can now make a digital copy of their own voice and use it to dub their videos. The technology behind this is called voice cloning, and you, as a YouTube creator, video editor, or a dubbing team, can use it to dub YouTube videos in your own voice. Read how in our new article 👇

Honored to have been part of the incredible Music China 2025 in Shanghai! It was a true privilege to contribute to this unique platform where Eastern and Western music technologies converge, creating new possibilities for the future of sound. Our Head of China Market, Anna Lyashchenko, was thrilled to connect with brilliant minds and explore how LALAL.AI can help shape the next chapter of music creation. It was great to catch up with: — Joya Hoya (禾急) from Positive Grid to discuss jamming technology; Xu Tao (徐滔) from WeiKing Acoustics to talk about the specifics of real-time vocal music separation; Richard Gu from Replica to explore possibilities of the LALAL.AI VST plugin. We also connected with innovators from: Yooba Technology, iCON, ACE studio, Antelope Audio, 奏浪APP, and A&C Pro Audio. The synergy from these meetings has given us incredible momentum to push the boundaries of audio AI even further. Catch you next time, Shanghai! ✋ 直到下次, 上海！

Quando apro eventi di grandi dimensioni come DJ, mi piace spesso iniziare con un brano in versione orchestrale. Questa volta ho voluto sperimentare con l’intelligenza artificiale per creare una versione orchestrale di “Man I Need” di Olivia Dean (se cercate un’idea regalo per Natale, questo album è una bella scoperta). **IL PROCESSO CHE HO SEGUITO:** Ho iniziato usando Claude per elaborare un documento di 47 pagine contenente tutte le istruzioni tecniche di Suno (l’AI generativa per contenuti musicali). Questo mi ha permesso di creare le basi per un prompt ottimale. Successivamente, ho creato un progetto su ChatGPT con queste istruzioni come fondamento. La richiesta era semplice: “realizza un prompt per Suno di una versione orchestrale con un crescendo di un paio di minuti”. Il risultato è stato questo: Orchestral intro. A two-minute emotional crescendo that starts as a breath — faint strings and airy pads shimmering like dawn light — and grows into a majestic cinematic overture. Layered violins and cellos in legato rise with warmth; brass enters gradually with bright resonance; timpani rolls, sub booms, and reversed cymbal swells add motion. Synth pads (Juno-60, Prophet-5) and angelic choirs blend with the orchestra, forming a luminous hybrid soundscape. Each 20 seconds, intensity builds — harp glissandos, choral bursts, and orchestral hits create a sense of awe and anticipation. The tempo accelerates naturally, mirroring a heartbeat shifting from calm to euphoric revelation. Emotion: wonder, elevation, positive tension. Mixing: hi-fi cinematic width, cathedral reverb on strings and choir, dynamic automation following the BPM rise. Goal: an awe-inspiring prelude that transforms silence into radiant energy before the main song begins. Il passo successivo è stato separare la strumentale dalla voce del brano originale usando LALAL.AI. Ho poi caricato la traccia strumentale su Suno insieme al prompt creato in precedenza, ottenendo esattamente la versione orchestrale che cercavo. L’ultimo step è stato l’esportazione e la sincronizzazione del prodotto di Suno con la voce originale su Ableton Live, dove ho rifinito i dettagli per ottenere il risultato finale. Lo trovate a questo link: https://lnkd.in/dA8xjxiH È stato un esperimento che mi ha fatto capire quanto l’AI possa diventare uno strumento creativo quando la guidi nel modo “giusto” Il risultato? Ascoltatelo e fatemi sapere che ne pensate! C I A O

💭 We were casually browsing Substack the other day and noticed that we were featured in some article as "consumer stem separation app". We found it cute, to be honest, because we branded ourselves as such when we just launched LALAL.AI. At first, we even thought that making karaoke backtracks was one of our primary use cases... Now, five years later though, our solutions are used by audio engineers, live recording engineers, music producers, videographers, SaaS, labels, and large teams in quite complicated scenarios we couldn't even think of when we started. Now, it's:  ✔️Almost 7 million users  ✔️13M+ hours of audio split  ✔️Featured in Billboard, Forbes, The Times, and many more outlets as well as influencers and celebrities More to go!

💬"Can LALAL.AI remove music from our influencer video and leave only their voice-over?" This is the question our business development team once got. Quick answer: you can! More than that, with LALAL.AI Voice Cleaner, you can: 🟡 Remove any music, like copyright tracks, accidental background music, or post-added sound, from any recording: live, recorded, or post-edited. 🟡 Make your content safe to publish anywhere without copyright strikes. 🟡 Scale it via API: integrate into SaaS platforms, video editors, or other tools for automatic processing. See how a video with music + voice transforms into clean, ready-to-publish voice (taken from our own influencer video) 👇

That's exactly the point: you can use several LALAL.AI products in one scenario, as in a bundle. Let's imagine a production team working on a documentary about an iconic music band from the 90s whose original live recordings are far from perfect: the vocals are muddy, the instruments overpower each other, and the sound echoes harshly in the big arena. Here's how LALAL.AI would work as a bundle for this project: 🟡 Sound engineer uses Stem Splitter to isolate vocals and instruments. 🟡 Mixing engineer applies Echo & Reverb Remover to clean stadium echo. 🟡 Audio editor runs Voice Cleaner on interviews and crowd noise. 🟡Localization specialist uses Voice Cloner to recreate band members’ voices in other languages. 🟡Producer / remixer applies Lead & Back Vocal Remover to polish tracks for the credits. Outcome: Clean, clear, and authentic audio ready for the documentary release. 🔗To try all of these, please visit https://www.lalal.ai/

💬"I haven't found anything that quite compares to LALAL.AI. I did try an iZotope algorithm once, but it didn't quite live up to my expectations." For Naples-based producer Raffaele Arcella, better known as Whodamanny, precision and emotion are two sides of the same coin. As one of the creative forces behind Periodica Records, he’s built a sound rooted in Italy’s analog heritage yet open to modern experimentation. In our new conversation with Raffaele, he shares how tools like LALAL.AI fit naturally into his analog-heavy workflow, helping isolate reference drums, fix mixes on the fly, and keep creativity flowing without ever compromising the soul of the sound.