Summary
Full Transcript
Join my FREE Skool community to learn about all things voice and voiceover: https://www.skool.com/voiceovermasterclass/about Discover Your Voiceover Potential and Turn Your New Skills into Exciting Opportunities as a Beginner in Voice Acting: https://www.voiceovermasterclass.com/voiceover-for-beginners Discover the secrets to breathing life into scripts and becoming the voice behind compelling narratives. Whether you dream of voicing radio or TV commercials, corporate video reports, conference announcements, documentary narrations, or even delving into the realms of video games and animations, this course is your ticket to voiceover success. REMEMBER: Without implementing what you learn, nothing will change! You can't improve your voice by just watching videos, informative as they are; you need to take action and practice and you'll soon see results! 📋 You can download our Free Voice Training PDF here: https://www.voiceovermasterclass.com/voice-training-pdf/ ✅ Remember to hit the subscribe button now and don’t forget to click the 🔔 bell icon and choose to receive all notifications, so that you don’t miss out on our latest videos that we upload weekly. https://www.youtube.com/voiceovermasterclass Join my FREE Skool community to learn about all things voice and voiceover: https://www.skool.com/voiceovermasterclass/about the biggest speech markets today are: 1. audiobooks 2. YouTube narration 3. corporate training 4. gaming dialogue 5. AI assistants. These sectors generate millions of minutes of audio per month. The longer the script, the more royalty you get! So, once you get accepted by an AI platform, what should you do to make sure that technically they’ll accept your data sets? Here are my top 10 tips! Here are The 10 Recording Mistakes That Ruin AI Voice Clones Even professional studios sometimes produce datasets that train poor-quality models. These are the most common causes. 1. Inconsistent microphone distance If the narrator moves toward or away from the mic, the model learns multiple timbres of the same voice. Result: • unstable tone • “phasey” AI speech. Fix Keep the same distance (usually 6–8 inches) throughout the entire dataset. 2. Recording multiple sessions with different mic setups If some recordings use different: • microphones • preamps • EQ settings the AI model learns two different voices. Fix Record the entire dataset using identical hardware and gain levels. 3. Excessive compression or EQ Many voice actors instinctively “polish” their recordings. This harms training because the AI learns processed audio rather than the natural voice. Fix Record clean, flat audio. Avoid: • compression • EQ • de-essing • noise reduction. 4. Room reflections Even mild reflections can confuse speech models. Result: • metallic or “hollow” AI voice. Fix Use: • acoustic panels • bass traps • a treated booth. 5. Background noise Noise floors above about -55 dB can degrade training data. Examples: • computer fans • street noise • HVAC systems. Fix Aim for -60 dB noise floor or lower. 6. Reading too theatrically Actors often perform too dramatically during dataset recording. This causes: • exaggerated patterns of rhythm and sound, stress and intonation • unnatural cadence. Fix Speak in neutral conversational tone unless you really are intending to clone a character voice. Keep the same voice style throughout. I suggest you do the whole data set in one session. 7. Inconsistent speaking speed If the narrator alternates between slow and fast delivery, the AI model produces erratic pacing. Fix Maintain consistent pace. 8. Mouth noises and clicks Lip smacks or saliva sounds get embedded in the model. Result: • audible clicks in generated speech. Fix Use: • hydration • apple slices (common trick) • light editing of clicks. In Adobe Audition there is an Automatic click remover…but use it with caution, you may delete some important syllables. 9. Incomplete phonetic coverage If the dataset doesn’t include enough examples of certain sounds, the AI struggles with pronunciation. Common missing sounds: • technical terms • numbers • acronyms. Fix Include balanced phonetic coverage. 10. THE MOST IMPORTANT TIP Poor transcript alignment Each recorded sentence must match its exact transcript. If there is a comma, you pause. You don’t add pauses where there is none stated in the script you’re given. You definitely don’t adlib. If there’s a contraction such as it’s, say it’s not “it is” and so on. If transcripts are wrong, the model learns incorrect pronunciation. Look, I hope all this is useful and if you would like to get more training on all things voiceover, whether you’re starting out – you can find my courses at Voiceover Masterclass.com, or join my free SKOOL community. The links are on screen now. Have a great day! Join my FREE Skool community to learn about all things voice and voiceover: https://www.skool.com/voiceovermasterclass/about
Continue this lesson in the app
Install CourseHive on Android or iOS to keep learning while you move.
