# Voice your client videos without re-recording: an ElevenLabs build with the consent rules Source: https://getstartedwithai.net/tutorials/elevenlabs-voice-cloning-ethics-setup Site: Get Started With AI Tagline: Build AI automations that actually run your business. Audience: solo professionals, freelancers, consultants and AI implementers in training Contact: hello@getstartedwithai.co.uk This Markdown companion is generated from the prerendered page for AI crawlers. # Voice your client videos without re-recording: an ElevenLabs build with the consent rules Three production use cases solved by one well-set-up ElevenLabs Professional Voice Clone: an explainer video voiceover that survives endless script revisions, a podcast translated into 32 languages with your voice intact, and an audio version of every blog post you publish. The catch is that voice cloning is the highest-risk feature in the AI stack. This build covers the consent and disclosure rules as carefully as the configuration itself, because the legal exposure under the EU AI Act and UK best practice is real, and the trust cost on a bad clone is permanent. ## Prep before you build (10 minutes) - An ElevenLabs account at elevenlabs.io. Creator from about £18 a month is the right starting plan. - A government-issued ID for verification. - A USB condenser mic (£100 to £200 mic produces noticeably better clones). - A treated room or quiet space to record the training set. ## The ethics rules: read these first Voice cloning is a tool that can be misused easily. These rules are non-negotiable. - You only clone a voice you own or have explicit, written permission to clone. Never a celebrity, public figure, client without explicit consent, or anyone who has not signed off. - Consent must be specific. "Clone my voice for this podcast translation" is fine. "Clone my voice for any future use" is not. - You disclose AI-generated audio in any public-facing context. The EU AI Act requires it; UK best practice expects it. - You verify the voice with ElevenLabs Identity Verification before doing any Professional clone. - Generated audio is watermarked. Do not strip the watermark; it is your defence in any dispute. ## Pricing in April 2026 - Free: 10,000 characters a month, three custom voices, no commercial license. - Starter at £4 a month: 30,000 characters, instant voice cloning, basic commercial license. - Creator at £18 a month annual (50 percent off first month promotion): 100,000 characters, professional voice cloning, full commercial license. - Pro at £80 a month: 500,000 characters, professional cloning, projects feature. - Most solo professionals start on Creator. The professional cloning is what makes voice work convincing. ## Step 1: Run identity verification (10 minutes) 1. Sign up at elevenlabs.io. Choose the Creator plan. 2. Open Verification. Upload a government-issued ID and complete the live capture (a 30-second video reading a phrase). 3. Verification approval takes 24 to 48 hours. Without verification, you cannot create a Professional Voice Clone. 4. Once approved, the badge appears on your profile. ## Step 2: Record the voice training set (30 minutes) Professional cloning needs about 30 minutes of clean audio. Quality matters more than quantity. - Use a USB condenser mic (£100 to £200 mic produces noticeably better clones). - Record in a treated room or a quiet space. Avoid echo, fans, traffic. - Read varied content: a news article, a personal story, a technical paragraph, a casual chat. Variety teaches the model your range. - Avoid reading flatly. Speak as you normally would, with natural pace and inflection. - Export as WAV or high-bitrate MP3. ## Step 3: build the Professional Voice Clone (15 minutes) 1. In ElevenLabs, click Voices, then Add a new voice, then Professional Voice Clone. 2. Confirm consent: tick the box that you own the voice or have explicit permission. 3. Upload the audio files. The system processes for 4 to 12 hours. 4. Once trained, test with the standard test phrases. The voice should match the sample on tone, pace and accent. 5. If the clone sounds wrong, the training audio was not varied enough. Re-record and retrain. ## Step 4: Ship the three real use cases (15 minutes) - Use case A: voice-over for client explainer videos. Record once, reuse the voice for any script. Saves studio time on every revision. - Use case B: podcast translation. Run the original episode through ElevenLabs Dubbing; the cloned voice speaks the translated version in 32 languages while keeping your voice characteristics. Useful for expanding podcast reach. - Use case C: long-form audio versions of blog posts. Run the post through TTS in your cloned voice. Adds a podcast-format channel without recording each episode. ## Step 5: Apply the disclosure templates (5 minutes) Use these in any public AI-generated audio. - Public podcast: "This episode includes a translated version recorded by AI using a clone of my voice with my consent. The original recording is also available." - Explainer video: "This video uses an AI-generated voice. The voice is mine, recorded once and reused with my consent." - Audio blog: "Audio version generated by AI using my own voice clone." ## What you ship A verified Professional Voice Clone of your own voice with documented consent. Three working use cases live: explainer voiceover, translated podcast, audio blog. About 4 hours of studio time replaced per script revision and a new audio channel for every blog post you publish. ## What to build next Pair this with the AI privacy and security tutorial for the broader operational hygiene, and the AI content pipeline tutorial for the audio channel that fits the full idea-to-publish flow. ## Related builds More tutorials in Content & Design. Each one runs in 60 to 90 minutes and ships a finished asset. ### Ship 5 client-ready assets without a designer: the Canva Magic Studio build 30 to 60 minutes. Solo Pro. [Open tutorial](/tutorials/canva-magic-studio-practitioner-2026) ### Cut design time per asset to 6 minutes: a 5-asset Canva AI run 60 to 90 minutes. Solo Pro. [Open tutorial](/tutorials/canva-ai-5-designs-30-min) ### Generate brand-consistent images without a designer: a Midjourney starter build 90 minutes. Solo Pro. [Open tutorial](/tutorials/midjourney-for-non-designers-2026) ### Cut video edits from days to hours: the AI video tools that survive client work 90 minutes. Solo Pro. [Open tutorial](/tutorials/ai-video-editing-tools-client-work) ## Questions people ask before they start ### Is voice cloning legal in the UK? Yes when you have explicit consent from the voice owner. The EU AI Act and UK best practice require disclosure on AI-generated audio in public contexts. Cloning a public figure or anyone without consent is illegal under existing privacy and image rights law. ### Can I detect if audio is AI? Detection is improving but not reliable. ElevenLabs adds a watermark to all generated audio that detection tools can read. Independent detection from raw audio alone is currently unreliable below 80 percent accuracy. ### How good is the clone? Professional Voice Clone with 30 minutes of clean training audio reaches about 95 percent likeness on your own voice. Friends and family identify the clone correctly about 60 percent of the time, but only when listening for it. ### What if a client asks me to clone someone elses voice without consent? Refuse. The risk is reputational and legal, and the brand promise of "AI that works" does not include "and gets you sued". Walk away from any engagement that requires it. ### How much should I charge to set up voice cloning for a client? Voice cloning for a clients own voice is a £600 to £1,800 setup. Identity verification, training session, clone build, three test outputs, plus a one-page consent and disclosure policy.