How to Record Voice Samples and Calibrate Custom AI Voice Clones in Lovo Genny

Creating a custom synthetic voice clone in Lovo.ai Genny required a straightforward yet precise two-phase process: capturing a clean acoustic voice sample of one to two minutes, followed by granular timeline calibration (adjusting pitch, tempo, custom pauses, and pronunciation) directly inside Genny's multi-track video editor. When executed properly, this workflow allowed marketers, educators, and content creators to generate scalable voiceovers without booking recurring studio recording sessions.

However, creators exploring this workflow today must account for a decisive industry development: in May 2026, Lovo, Inc. filed for Chapter 7 bankruptcy liquidation following landmark voice-cloning litigation (Lehrman & Sage v. Lovo, Inc. in the U.S. District Court for the Southern District of New York), leading to the complete shutdown and inaccessibility of the lovo.ai platform. Understanding how Genny structured voice recording and calibration provides valuable technical principles for synthetic voice production, while also illuminating the critical consent standards and active alternative tools creators must rely on today.

1. Preparing and Recording the Voice Sample

The accuracy and naturalness of any neural voice clone depend fundamentally on the quality of the training audio. Genny's neural voice engine analyzed vocal timbre, cadence, pitch range, and breathing patterns from user uploads. Any room echo, background hum, or erratic vocal delivery fed into the system became permanently embedded in the resulting voice model.

Acoustic Environment and Setup

• Acoustic Treatment: Record in an acoustically dead, quiet room. Avoid rooms with hard, reflective surfaces such as bare walls, tile, or large windows. Small rooms with carpets, heavy drapes, or closets filled with clothing provide effective natural sound absorption that eliminates flutter echo.
• Microphone Selection: Use a dedicated cardioid condenser or dynamic broadcast microphone connected via USB or an XLR audio interface. Built-in laptop microphones, smartphone speakerphones, and Bluetooth earbuds introduce digital compression artifacts, room reverberation, and frequency clipping that degrade voice synthesis.
• Microphone Technique: Maintain a consistent distance of 6 to 8 inches (15 to 20 cm) from the capsule. Always install a pop filter or foam windscreen to disperse plosive bursts (harsh "p", "b", and "t" sounds). Position the microphone at a slight 45-degree angle toward the mouth rather than directly head-on to reduce sibilance and breathing noise.
• Gain and Volume Levels: Calibrate microphone gain so the speech peaks safely between -6 dB and -3 dB, with an average RMS volume between -23 dB and -18 dB. Avoid digital clipping (hitting 0 dB), which creates permanent distortion that corrupts AI training models.

Sample Runtime and Vocal Performance

• Duration: Genny required a minimum of one minute of clean audio, with one to two minutes representing the optimal target for instant voice cloning. Submitting more than three minutes to rapid cloning models yielded diminishing returns and could introduce conflicting tonal variations.
• Vocal Consistency: Speak in the exact tone, pacing, and emotional register intended for final production. If your project demands an authoritative corporate explainer voice, maintain steady, confident pacing. If the clone is intended for conversational video podcasts, maintain a relaxed, engaging cadence. Avoid alternating between whispered speech, shouting, or extreme pitch changes within the same sample.
• Reading Script: Read a phonetically rich script containing complete sentences, varied punctuation (questions, declarative statements), and standard phonemes. Avoid long silences, throat clearing, paper shuffling, or mouth clicks.

2. Ingesting and Generating the Clone in Genny

Once audio samples were recorded and trimmed, creators imported them into Genny's web interface to build the synthetic voice model.

1. File Specifications: Export recordings as clean, uncompressed WAV files (16-bit or 24-bit at 44.1 kHz or 48 kHz) or high-bitrate MP3 files (192 kbps or higher). Ensure all background music, ambient beds, and secondary voices are entirely removed.
2. Accessing the Voice Cloning Module: Log in to the Genny workspace, navigate to the voice selector panel or the left sidebar menu, and select Voice Cloning (or Create Custom Voice).
3. Audio Ingestion: Choose between uploading pre-recorded audio files or recording directly into the browser via Genny's integrated recorder. Uploading clean, pre-edited studio files consistently produced superior results compared to raw in-browser captures.
4. Metadata Configuration: Assign a descriptive name to the cloned voice profile, specify the speaker's gender, and select the primary language and accent tags (such as English - US or English - UK) to help Genny map phonetic models accurately.
5. Rights and Consent Confirmation: Complete the platform's verification check, confirming that you possess full commercial rights and explicit consent from the voice owner.
6. Processing: Submit the audio. Genny's neural voice engine synthesized the voice characteristics, generating an interactive voice clone profile in your personal voice library within one to three minutes.

3. Calibrating and Fine-Tuning the Voice Model on the Timeline

A raw voice clone rarely sounds perfect straight out of automated synthesis. Genny provided a suite of timeline-level calibration tools designed to polish synthetic speech into broadcast-ready voiceovers.

Pitch and Speed Adjustments

• Pitch Shifting: In Genny's voice settings sidebar, adjust the global or block-specific pitch slider. If a cloned voice sounds overly resonant or artificially boomy, nudge the pitch up by 2% to 5% to restore crispness. Conversely, lowering pitch slightly can add gravitas to narrative explainers.
• Pacing and Tempo: Modulate speed between 0.5x and 2.0x. Fast conversational pacing often requires setting speech rate around 1.05x to 1.1x to prevent sluggish delivery, whereas instructional tutorials benefit from a standard 1.0x or 0.95x rate to preserve comprehension.

Custom Pauses and Punctuation Dynamics

• Pause Insertion: Synthetic voices often rush between clauses, creating an unnatural, breathless delivery. Genny allowed creators to place custom pause markers directly between words or sentences, specifying exact durations (e.g., 200ms for a comma pause, 500ms to 800ms for a full sentence break).
• Punctuation Tuning: Punctuation acts as an acoustic guide for AI models. Replacing standard periods with ellipses (...) softens terminal sentence drops, while em-dashes (—) introduce conversational mid-sentence pauses.

Pronunciation Editor

• Specialized terminology, proprietary product names, acronyms, and non-English loanwords often confound text-to-speech phonetic engines.
• Using Genny's Pronunciation Editor, creators could create phonetic substitution dictionaries (e.g., replacing "SaaS" with "sass" or specifying custom phonetic spellings) applied globally across the project to ensure flawless articulation.

Multi-Track Timeline Synchronization

• In Genny's multi-track editor, assign the cloned voice track across segmented text blocks.
• Align voiceover clips directly with background audio tracks, video footage, sound effects, and auto-generated captions, utilizing volume ducking so background audio automatically drops during voice narration.

4. The Shutdown of Lovo.ai and the Lehrman v. Lovo Precedent

While Genny's voice cloning interface offered a pioneering all-in-one content creation workflow, the platform is no longer operational. Creators seeking to deploy this workflow must understand the legal and operational factors behind its demise.

The Bankruptcy Liquidation

On May 27, 2026, Lovo, Inc. filed for Chapter 7 bankruptcy liquidation in the U.S. Bankruptcy Court for the Southern District of New York (Case No. 26-11249). Under Chapter 7 liquidation, the company ceased all business operations, its technical infrastructure was decommissioned, and the official lovo.ai web deployment went offline. Accounts, projects, and custom voice models hosted on Genny became permanently unavailable.

The Lehrman & Sage v. Lovo, Inc. Lawsuit

The operational collapse followed intense legal scrutiny stemming from a class-action lawsuit filed by professional voice actors Paul Skye Lehrman and Linnea Sage (Lehrman & Sage v. Lovo, Inc., Case No. 24-CV-3770, S.D.N.Y.). The plaintiffs alleged that Lovo recruited them via freelance platforms under the pretext of recording voice samples for "internal academic research," only to repurpose those recordings to train Genny's commercial voice generator and sell unauthorized synthetic voice clones (marketed under names such as "Kyle Snow" and "Sally Coleman").

In a milestone July 2025 ruling, the federal court addressed key boundaries of AI voice cloning:

• Copyright Limitations: The court held that federal copyright law protects fixed sound recordings rather than an abstract vocal timbre or style, dismissing certain copyright infringement and federal trademark claims.
• Right of Publicity and Contract Protections: Crucially, the court permitted claims for breach of contract and violations of New York Civil Rights Law §§ 50 and 51 (right of publicity) to proceed. The court recognized that using an individual's voice to train an AI model and repeatedly generate synthetic outputs constitutes an unauthorized commercial use of their personal identity.
• Procedural Stay: Following Lovo's Chapter 7 bankruptcy filing in mid-2026, the litigation was automatically stayed, leaving creators and commercial users with a clear lesson in legal liability and platform risk.

Core Takeaways for Commercial Creators

1. Never Clone Without Documented Consent: Utilizing voice samples without explicit, written commercial licensing agreements exposes both the platform and the end user to right-of-publicity and breach-of-contract liabilities.
2. Back Up Audio Assets Locally: Never rely solely on cloud platforms to preserve custom voice assets. Maintain raw WAV recordings and text transcripts in local, version-controlled storage.

5. Active Commercial Alternatives for AI Voice Cloning

Because Lovo Genny is defunct, content creators requiring high-fidelity voice cloning should migrate their workflows to established, legally compliant platforms:

By adhering to rigorous recording standards—quiet acoustic environments, quality cardioid microphones, uniform vocal energy, and careful post-generation calibration—creators can achieve broadcast-caliber voice clones across modern platforms while ensuring full legal and contractual compliance.

References

• New York Court Tackles the Legality of AI Voice Cloning | Insights | Skadden, Arps, Slate, Meagher & Flom LLP
• Instant Voice Cloning | ElevenLabs Documentation
• Lehrman v. Lovo Inc. | Loeb & Loeb LLP