{"course":{"slug":"choosing-your-tools","title":"Choosing and Using the Tools","url":"https://addaly.com/learn/choosing-your-tools","level":"beginner","track":"work"},"lesson":{"slug":"voice-and-speech-tools","title":"Voice and speech tools","url":"https://addaly.com/learn/choosing-your-tools/voice-and-speech-tools","position":59,"of":76,"minutes":8,"answer":"Text-to-speech is free, local and good enough for accessibility, proofreading and reaching people who do not read the language.","keyIdea":"Text-to-speech is free, local and good enough for accessibility, proofreading and reaching people who do not read the language, while voice cloning carries a real fraud risk that makes explicit consent and a household callback rule the load-bearing parts."},"body":{"format":"markdown","text":"## What the category contains\n\n**Text to speech.** Turning writing into audio. Free and built into every operating system; better voices available from paid services and from open models.\n\n**Voice cloning.** Producing speech in a particular person's voice from a sample, sometimes only seconds long.\n\n**Speech to speech.** Changing an existing recording to a different voice, keeping the timing and delivery.\n\n**Dubbing.** Translating a recording and speaking it in the new language, sometimes matching the original speaker's voice.\n\nThe technology has moved fast, and the honest summary is that the good hosted services are now hard to distinguish from a person on a short clip, and the free options are good enough for most practical uses.\n\n## The free path\n\n**Piper** runs locally, produces natural speech in many languages, and is fast enough on a Raspberry Pi. This is the workhorse for accessibility, for reading documents aloud, and for anything you produce regularly.\n\n**Coqui TTS** and similar projects offer more voices and cloning at the cost of more setup.\n\n**Your operating system.** Built-in voices have improved considerably. For reading a document back to yourself, no installation is required at all.\n\n## Where these help\n\n- **Accessibility.** For someone who reads with difficulty or not at all, having any document read aloud in a natural voice is transformative, and it is free.\n- **Proofreading.** Listening to your own writing exposes errors the eye slides over. This is an old editor's trick and it works better with a good voice.\n- **Learning a language.** Hearing a sentence read correctly, on demand, at your own pace.\n- **Reaching people who do not read your written language** but speak it: a large population in many countries, and one that written-only material excludes entirely.\n- **Producing audio versions** of material at a cost that makes it worth doing for small audiences.\n\n## The consent problem, which is the important part\n\nVoice cloning creates a genuine harm that has already happened many times. Recorded voices are used in fraud: the call that sounds like a family member in trouble, the instruction that sounds like a manager approving a payment. A few seconds of audio from a public video is enough for some tools.\n\nThree things follow.\n\n**Do not clone a voice without the person's explicit permission.** Reputable services require an attestation, some require a spoken consent recording, and it remains the user's responsibility. Several jurisdictions now have laws about voice likeness, and the rules differ.\n\n**Assume other people's voices can be cloned.** The practical defence for your family and your workplace is a shared phrase or a callback rule for anything involving money or urgency. That advice sounds excessive until it has happened to someone you know.\n\n**Disclose synthetic voice where the listener would care.** In advertising, in journalism, in anything a person might rely on. Some platforms and some regimes now require labelling, and the requirements are still forming.\n\n## Pronunciation, which is where most of the work goes\n\nThe gap between an acceptable read and a good one is almost entirely names, acronyms and foreign words. Every serious tool accepts a pronunciation override: a phonetic spelling, a dictionary of terms, or SSML markup that specifies how a word is said.\n\nBuilding that dictionary once, for the twenty words your material keeps using, is an hour that improves everything you produce afterwards. It is also the reason a free local tool with a good dictionary often beats an expensive service used straight out of the box, and it is the part of the work no product page mentions.\n\n## Choosing a service\n\nBeyond quality, three things separate them for real work:\n\n- **Languages and accents**, particularly whether your language is supported at all rather than as an accented approximation.\n- **Control over delivery**: pauses, emphasis, pace, pronunciation of names. On a long piece this is what turns an acceptable read into a good one.\n- **Commercial terms.** Whether you may use the audio commercially, whether the voice is licensed for it, and what happens to the voice you cloned. Read this before recording a client's audiobook, not after.\n\nFor anything you produce regularly, test the free local option first. The gap between Piper and a paid service is real and it is smaller than the gap between doing this at all and not doing it."},"check":{"question":"A company wants to use a cloned executive voice for internal announcements. Beyond audio quality, what is the most important consideration?","options":[{"text":"Whether the audio files are stored securely, since a leaked recording could be reused elsewhere.","correct":false,"tempting":"Protects the outputs while leaving the underlying capability (a voice model of a named executive) unaddressed."},{"text":"Whether staff can tell the difference, since an announcement that sounds real but is not may reduce trust.","correct":false,"tempting":"Raises a presentational concern that matters less than establishing consent and a way to verify genuine instructions."},{"text":"Whether the service supports the accents and pacing needed for the executive's usual delivery.","correct":false,"tempting":"Optimises for fidelity, which increases rather than reduces the risk that a fraudulent message will be believed."},{"text":"Explicit consent, and a verification rule so staff can tell a real instruction from a cloned one.","correct":true}]},"prev":{"slug":"transcription-and-translation","title":"Transcription and translation","url":"https://addaly.com/learn/choosing-your-tools/transcription-and-translation"},"next":{"slug":"video-tools","title":"Video tools, generated and edited","url":"https://addaly.com/learn/choosing-your-tools/video-tools"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Voice and speech tools', in Choosing and Using the Tools. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/choosing-your-tools/voice-and-speech-tools (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Voice and speech tools","courseTitle":"Choosing and Using the Tools","module":{"index":7,"slug":"beyond-chat","title":"Beyond the chat box"},"answer":null,"keyIdea":"Text-to-speech is free, local and good enough for accessibility, proofreading and reaching people who do not read the language, while voice cloning carries a real fraud risk that makes explicit consent and a household callback rule the load-bearing parts.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"what-the-category-contains","runs":[{"text":"What the category contains"}]},{"t":"p","runs":[{"text":"Text to speech.","b":true},{"text":" Turning writing into audio. Free and built into every operating system; better voices available from paid services and from open models."}]},{"t":"p","runs":[{"text":"Voice cloning.","b":true},{"text":" Producing speech in a particular person’s voice from a sample, sometimes only seconds long."}]},{"t":"p","runs":[{"text":"Speech to speech.","b":true},{"text":" Changing an existing recording to a different voice, keeping the timing and delivery."}]},{"t":"p","runs":[{"text":"Dubbing.","b":true},{"text":" Translating a recording and speaking it in the new language, sometimes matching the original speaker’s voice."}]},{"t":"p","runs":[{"text":"The technology has moved fast, and the honest summary is that the good hosted services are now hard to distinguish from a person on a short clip, and the free options are good enough for most practical uses."}]},{"t":"h2","id":"the-free-path","runs":[{"text":"The free path"}],"role":"free"},{"t":"p","runs":[{"text":"Piper","b":true},{"text":" runs locally, produces natural speech in many languages, and is fast enough on a Raspberry Pi. This is the workhorse for accessibility, for reading documents aloud, and for anything you produce regularly."}]},{"t":"p","runs":[{"text":"Coqui TTS","b":true},{"text":" and similar projects offer more voices and cloning at the cost of more setup."}]},{"t":"p","runs":[{"text":"Your operating system.","b":true},{"text":" Built-in voices have improved considerably. For reading a document back to yourself, no installation is required at all."}]},{"t":"h2","id":"where-these-help","runs":[{"text":"Where these help"}]},{"t":"ul","items":[[{"text":"Accessibility.","b":true},{"text":" For someone who reads with difficulty or not at all, having any document read aloud in a natural voice is transformative, and it is free."}],[{"text":"Proofreading.","b":true},{"text":" Listening to your own writing exposes errors the eye slides over. This is an old editor’s trick and it works better with a good voice."}],[{"text":"Learning a language.","b":true},{"text":" Hearing a sentence read correctly, on demand, at your own pace."}],[{"text":"Reaching people who do not read your written language","b":true},{"text":" but speak it: a large population in many countries, and one that written-only material excludes entirely."}],[{"text":"Producing audio versions","b":true},{"text":" of material at a cost that makes it worth doing for small audiences."}]]},{"t":"h2","id":"the-consent-problem-which-is-the-important-part","runs":[{"text":"The consent problem, which is the important part"}]},{"t":"p","runs":[{"text":"Voice cloning creates a genuine harm that has already happened many times. Recorded voices are used in fraud: the call that sounds like a family member in trouble, the instruction that sounds like a manager approving a payment. A few seconds of audio from a public video is enough for some tools."}]},{"t":"p","runs":[{"text":"Three things follow."}]},{"t":"p","runs":[{"text":"Do not clone a voice without the person’s explicit permission.","b":true},{"text":" Reputable services require an attestation, some require a spoken consent recording, and it remains the user’s responsibility. Several jurisdictions now have laws about voice likeness, and the rules differ."}]},{"t":"p","runs":[{"text":"Assume other people’s voices can be cloned.","b":true},{"text":" The practical defence for your family and your workplace is a shared phrase or a callback rule for anything involving money or urgency. That advice sounds excessive until it has happened to someone you know."}]},{"t":"p","runs":[{"text":"Disclose synthetic voice where the listener would care.","b":true},{"text":" In advertising, in journalism, in anything a person might rely on. Some platforms and some regimes now require labelling, and the requirements are still forming."}]},{"t":"h2","id":"pronunciation-which-is-where-most-of-the-work-goes","runs":[{"text":"Pronunciation, which is where most of the work goes"}]},{"t":"p","runs":[{"text":"The gap between an acceptable read and a good one is almost entirely names, acronyms and foreign words. Every serious tool accepts a pronunciation override: a phonetic spelling, a dictionary of terms, or SSML markup that specifies how a word is said."}]},{"t":"p","runs":[{"text":"Building that dictionary once, for the twenty words your material keeps using, is an hour that improves everything you produce afterwards. It is also the reason a free local tool with a good dictionary often beats an expensive service used straight out of the box, and it is the part of the work no product page mentions."}]},{"t":"h2","id":"choosing-a-service","runs":[{"text":"Choosing a service"}]},{"t":"p","runs":[{"text":"Beyond quality, three things separate them for real work:"}]},{"t":"ul","items":[[{"text":"Languages and accents","b":true},{"text":", particularly whether your language is supported at all rather than as an accented approximation."}],[{"text":"Control over delivery","b":true},{"text":": pauses, emphasis, pace, pronunciation of names. On a long piece this is what turns an acceptable read into a good one."}],[{"text":"Commercial terms.","b":true},{"text":" Whether you may use the audio commercially, whether the voice is licensed for it, and what happens to the voice you cloned. Read this before recording a client’s audiobook, not after."}]]},{"t":"p","runs":[{"text":"For anything you produce regularly, test the free local option first. The gap between Piper and a paid service is real and it is smaller than the gap between doing this at all and not doing it."}]}],"check":{"question":[{"text":"A company wants to use a cloned executive voice for internal announcements. Beyond audio quality, what is the most important consideration?"}],"options":[{"text":[{"text":"Whether the audio files are stored securely, since a leaked recording could be reused elsewhere."}],"correct":false,"tempting":[{"text":"Protects the outputs while leaving the underlying capability (a voice model of a named executive) unaddressed."}]},{"text":[{"text":"Whether staff can tell the difference, since an announcement that sounds real but is not may reduce trust."}],"correct":false,"tempting":[{"text":"Raises a presentational concern that matters less than establishing consent and a way to verify genuine instructions."}]},{"text":[{"text":"Whether the service supports the accents and pacing needed for the executive’s usual delivery."}],"correct":false,"tempting":[{"text":"Optimises for fidelity, which increases rather than reduces the risk that a fraudulent message will be believed."}]},{"text":[{"text":"Explicit consent, and a verification rule so staff can tell a real instruction from a cloned one."}],"correct":true,"tempting":null}],"order":[0,2,1,3],"why":null},"prev":{"slug":"transcription-and-translation","title":"Transcription and translation","minutes":8},"next":{"slug":"video-tools","title":"Video tools, generated and edited","minutes":8}}}