AI Voice with ElevenLabs

The voice layer for screen recordings, B-roll, and any content where you want your own voice without recording every time

ElevenLabs is the voice layer for content where you do not have a talking head. Screen recordings, B-roll compilations, slide-based demos: all of these become full UGC videos when you add a natural-sounding voice-over.

It is also the tool you use when you want your own voice on content without recording yourself each time. You clone your voice once and ElevenLabs generates narration in your voice from any text you provide.

Plans and Limits

The Free tier gives you 10,000 characters per month with no commercial rights and no voice cloning. This is sufficient for testing the quality before committing to a paid plan.

The Starter plan at $5 per month gives you 30,000 characters per month, commercial rights included, and Instant Voice Cloning available using a 3-minute audio sample. This is the plan to start with.

The Creator plan at $22 per month gives you 100,000 characters per month and Professional Voice Cloning which requires 30 minutes of training audio and produces a significantly higher quality clone. Worth moving to when you are producing video at scale.

For reference on character counts: 1,000 characters is approximately 150 words and 70 seconds of narration. The Starter plan's 30,000 characters is approximately 4,500 words and 35 minutes of audio per month. At 60 seconds per video the Starter plan gives you approximately 35 videos monthly.

When to Use ElevenLabs vs HeyGen

Use ElevenLabs when the content is a screen recording and no face is needed, when you want your own voice on content you are not filming, when you need voice for B-roll footage, or when you are producing audio content such as podcast clips or audio ads.

Use HeyGen when you need a talking head with a face visible, when you are creating testimonial-style content, or when the face-to-camera connection is part of the UGC format.

Use both together by using HeyGen for the avatar sequence and ElevenLabs for any sections where your own voice narrates over B-roll with CapCut handling the assembly.

Setting Up Your Voice Clone

Voice cloning on ElevenLabs starts at the Starter plan.

For Instant Voice Cloning go to Voices then Add a New Voice then Instant Voice Cloning. Record or upload 3 to 10 minutes of clean audio. The audio must be you speaking clearly with no music, no multiple speakers, and minimal background noise. A phone recording in a quiet room works. Studio quality is not required. Name the voice and save it.

What makes a good training sample: speak at your natural pace and not slower or more carefully than usual, include natural pauses and some variation in energy, read aloud from any text such as a book or an article or your own lesson notes. Avoid music in the background, people talking nearby, and air conditioning or fan noise.

After training generate a short test clip and compare it to your real voice. If the intonation patterns are off re-record with a longer sample.

Generating Voice-Over from a Script

Once your voice is set up follow these steps.

Go to Text to Speech in ElevenLabs. Select your cloned voice or any ElevenLabs voice if you have not cloned yours yet. Paste your script. Adjust the settings: set Stability at 60 to 70% for conversational narration where higher means more consistent and less expressive, set Clarity at 75% where higher is cleaner but slightly more robotic, and use Style sparingly where at 0 it uses the training audio's natural style. Preview before downloading. Download as MP3 or WAV.

Syncing Voice-Over to Screen Recording

This is the workflow for B-roll or screen recording content.

Step 1: Record your screen first

Screen record the system, website, or process you are demonstrating. Keep it raw and do not worry about perfect timing. You will sync the voice to the video in CapCut. Record slightly longer than you need because it is easier to trim footage than to fill gaps.

Step 2: Write the script to match the footage

Watch the screen recording. Write a script that describes what is happening on screen. Time it to fit within 30 to 45 seconds of narration for a 45-second video.

Use this prompt:

Using @UGC.md — write a 45-second voice-over narration script for a

screen recording that shows [describe what the recording shows].

The narration should:

• Open with a hook that explains why this matters and not "In this video

I'll show you"

• Describe what the viewer is seeing while they're seeing it
• Note key moments such as "Here's where the client saves the most time"
• End with a soft CTA

Write in conversational language as if walking a friend through this.

Step 3: Generate the voice in ElevenLabs

Paste the script, generate the audio, and download the file.

Step 4: Assemble in CapCut

Add the screen recording video to the CapCut timeline. Add the ElevenLabs audio on a separate track. Adjust the video speed or trim clips to match the audio pacing. CapCut's auto-sync is not reliable for this so manual alignment by ear is faster.

Using Stock ElevenLabs Voices

If you are not cloning your own voice ElevenLabs has a library of licensed voices.

For UGC content choose voices labelled Conversational or Narrative over Broadcast or Professional. For British English, Lily, George, and Daniel are consistently natural-sounding. For American English, Rachel and Adam work well. Preview several before deciding because the same text sounds markedly different across voices.

Avoid voices with excessive expressiveness for short-form social content. Conversational monotone with slight variation sounds more authentic than over-performed narration.

Audio Quality Settings for Export

When downloading from ElevenLabs use MP3 for social media or WAV for any platform that re-encodes at low bitrate. Set the sample rate to 44.1kHz and use a minimum bitrate of 128kbps.

In CapCut ensure the audio track is not clipping by checking for red peaks in the waveform. The typical ElevenLabs output is well-levelled so lower it slightly if you are adding background music.