AI Audio — ElevenLabs, Suno, and synthetic voices

Section 10.5.5 : AI Audio — ElevenLabs, Suno, and synthetic voices

🎯 Learning objective

Master AI audio tools — voice synthesis, voice cloning, music generation — to create podcasts, narrations, and professional audio content. You'll know how to choose the right tool, configure voice parameters, and integrate AI audio into your content creation workflows.

Synthetic voice becomes indistinguishable

In 2025-2026, AI-generated voices are virtually indistinguishable from human voices. A blind test would show that most people can no longer tell the difference from a natural voice. This opens immense possibilities for professional audio content — but also important ethical questions (see ethics section).

ElevenLabs: the voice synthesis reference

ElevenLabs is the undisputed leader in high-quality voice synthesis. Used by podcasters, e-learning creators, video game studios, and companies to automate audio content production.

Main features:

Generation parameters:

ElevenLabs workflow for an e-learning module:

1. →Write the script with ChatGPT (structure: introduction, content, summary, transition)
2. →Break into 2-3 minute sections (pay attention to natural pauses)
3. →Choose an appropriate voice (male/female, dynamic/calm, young/mature)
4. →Generate each section separately (easier to adjust)
5. →Add pauses: <break time="1s" /> in the text (SSML syntax)
6. →Listen and adjust: stability, clarity, pronunciation of technical terms
7. →Export as 320kbps MP3 for final quality

Pronunciation tip: for mispronounced technical terms, use phonetic pronunciation: <phoneme alphabet="ipa" ph="ˈpɹɒmpt">prompt</phoneme> or simply write the word as it's pronounced ("prommpt" instead of "prompt" if the French voice mispronounces it).

Pricing:

• →Free: 10,000 characters/month (~10 minutes of voice)
• →Starter ($5/month): 30,000 characters/month + 3 cloned voices
• →Creator ($22/month): 100,000 characters/month + unlimited voices + long projects

Voice Cloning: clone your own voice

ElevenLabs' voice cloning is remarkably faithful with only 5 minutes of audio sample:

How to prepare your sample:

1. →Record yourself in a quiet environment (no reverberation)
2. →Speak naturally: read varied text (not monotone), include questions, exclamations
3. →Use a decent mic (even AirPods work, but a USB mic is better)
4. →Duration: 5 minutes minimum. Longer = better clone (ideal: 15-30 minutes)
5. →Format: MP3, WAV, or M4A

What the clone can do:

• →Speak in 29+ languages (with your timbre and intonation, even in languages you don't speak)
• →Read any text with your voice
• →Maintain your vocal style (tone, rhythm, expressiveness)

Suno: complete music generation

Suno generates complete songs (vocals + instruments + arrangement) from a simple text description. This isn't generic royalty-free music — it's original music with lyrics, singing, and radio-ready production.

How to prompt Suno effectively:

Genre: acoustic indie pop
Tempo: moderate (110 BPM)
Mood: energetic and optimistic
Voice: female, clear, with some grit
Theme: digital transformation and continuous learning
Language: English
Structure: verse-chorus-verse-chorus-bridge-final chorus
Instruments: acoustic guitar, bass, light drums, piano
The chorus should be memorable and repetitive.

Concrete use cases:

• →Podcast jingle: 15-30 second intro/outro music, consistent with the podcast's identity
• →Video background music: 2-3 minute instrumental track matched to the video's mood
• →Hold music: original phone hold music
• →Event: theme music for a conference or product launch

Pricing: free (10 songs/day, non-commercial use) → $10/month (Pro, 500 songs/month, commercial use)

Udio: the high-fidelity music alternative

Udio is Suno's main competitor, with different strengths:

• →More "professional" music production (mixing, mastering)
• →Better rendering of classical genres (jazz, classical, rock)
• →Finer control of musical structure
• →Remixing from existing melodies

Suno vs Udio: Suno is more accessible and generates catchy "pop" songs. Udio excels in demanding musical genres and production quality.

NotebookLM: the automatic podcast

Google NotebookLM (free) generates automatic conversational podcasts — two virtual hosts naturally discuss your sources:

1. →Upload your sources: PDF, web articles, YouTube videos, Google Docs
2. →Click "Audio Overview"
3. →In 3-5 minutes, get a 10-20 minute natural dialogue

The result is impressive: the two hosts joke around, ask each other questions, simplify complex concepts, and maintain an engaging pace. It's the fastest way to transform written content into audio format.

Use cases:

• →Transform a 50-page report into a 15-minute podcast for your team
• →Create an audio version of a blog post
• →Summarize a series of academic papers into an accessible discussion
• →Prepare an audio briefing on a complex topic

AI audio tools summary table