Remember me
by Alexandra ThompsonTodayLocation audio does not always cooperate. There is wind, traffic, background music, room noise, or a failed primary microphone that can leave an editor with usable footage but dialogue that’s difficult to salvage. Addressing these common production problems, 2026 People’s Voice Webby Award winner LALAL.AI has introduced Lynx, a new AI model designed to separate speech from unwanted sounds in recordings. Let’s take a look.
CineD previously examined LALAL.AI in the broader context of the many AI tools available for audio post-production. The platform includes several tools for stem separation and voice processing, and Lynx now extends that lineup as the model behind its Voice Cleaner and Voice & Noise tools. It is designed specifically to remove noise and background music from recorded speech.
LALAL.AI select stems. Image credit: LALAL.AILynx and the Voice CleanerLynx is LALAL.AI’s first neural-network model developed exclusively for speech denoising. It does not replace the company’s broader stem-separation models or other voice-processing tools. Instead, it powers the existing Voice Cleaner and Voice & Noise stem, separating spoken audio from unwanted sounds.
Of particular note, Lynx can remove music captured at public events and other locations while preserving the spoken dialogue – useful when otherwise usable footage contains copyrighted music that cannot remain in the finished edit or could trigger a platform claim.
AI vocal remover app and plugins. Image credit: LALAL.AIBuilt for production dialogueOther possible uses include cleaning up interviews recorded near construction or traffic, outdoor conversations affected by wind, and backup recordings used after a primary microphone fails. According to LALAL.AI, Lynx is six times smaller than its Andromeda model and is designed specifically for speech denoising, separating speech from environmental noise and background music.
For context, Lynx is not the only tool designed to recover dialogue from difficult recordings. Adobe Podcast Enhance Speech can remove background noise, reverb, chatter, and music. iZotope RX offers a broader set of audio restoration tools, including Dialogue Isolate, spectral editing, and Scene Rebalance. Depending on your needs, Lynx may be the more accessible option, with a simpler workflow and a considerably lower initial cost, while RX provides more extensive repair tools and is available as a perpetual license.
Price and availabilityLynx is available now through LALAL.AI’s Voice Cleaner and as the model powering the Voice and Noise stem in its Vocal Remover and Stem Splitter tools. It can be accessed through a web browser, the mobile app, the desktop application in Cloud mode, and the LALAL.AI API. The company says a planned update will improve speech extraction when the speaker is farther from the microphone, which may be relevant to documentary and run-and-gun production.
LALAL.AI offers a number of subscription options, including a free Starter plan to preview results, but processed files cannot be downloaded. U.S. pricing starts at $9.99 per month for the Lite plan and $19.99 per month for the Pro plan, with lower monthly rates when billed annually. The company is also offering the annual Pro plan for $126 until July 28, 2026. For more information, please visit the LALAL.AI website
Do you currently use AI-based tools to repair location dialogue? How well have they handled recordings containing both speech and background music? Let us know in the comments.