Key takeaways
- Canva adds text to speech through marketplace apps, and CapCut uses AI neural voices.
- Leading speech to text models include Whisper, Parakeet, Canary, wav2vec 2.0 and commercial APIs.
- Speech to text is AI, and modern models generate text like generative models, so review outputs.
- Mac, Outlook and Notes all support dictation, with Windows key plus H as a universal option.
- Text-to-speech tracking highlights words as they are read, helping readers follow along.
Speech technology now appears in design tools, video editors, email and note apps, powered by neural models that turn text into voices and voices into text. Knowing which tools offer it and how the models work helps you use them well. I built the OCR & Speech Workspace with live speech-to-text, so this page covers Canva and CapCut voices, speech-to-text models, whether they count as AI or generative AI, dictation on Mac, Outlook and Notes, uses of text-to-speech and word tracking.
Does Canva have text to speech?
Yes, through apps in Canva's Apps marketplace. Voice apps let you type a script, choose an AI voice and add the generated voiceover to videos and presentations directly in the editor. Features and free usage vary by app and plan.
Search the Apps panel for text to speech or voiceover to see current options, such as voice generators from third-party providers. After generating audio, trim and position it on the timeline like any audio clip. Check licensing for commercial use, especially for client work or ads. For more control over pronunciation and pacing, generate voiceovers in a dedicated text-to-speech tool and upload the audio file to Canva.
What are speech to text models?
Speech-to-text models are neural networks that convert audio into text. Popular examples include OpenAI's Whisper, NVIDIA's Parakeet and Canary, Meta's wav2vec 2.0 family, and commercial models from Deepgram, AssemblyAI, Google and Microsoft.
Architectures differ: Whisper is an encoder-decoder transformer trained on huge multilingual audio datasets, Conformer models combine convolution and attention for efficient streaming, and wav2vec 2.0 learns from unlabelled audio before fine-tuning. Hugging Face's Open ASR Leaderboard compares open models on word error rate and speed. Choose by language support, accuracy on your audio, streaming needs and whether you need local deployment for privacy. Smaller models can run on phones and laptops for offline transcription.
Is speech to text generative AI?
Partly. Modern speech-to-text models such as Whisper generate text token by token, using the same transformer techniques as generative language models, so they are often grouped with generative AI. But their task is recognition, not creating new content.
The distinction matters for risk: generative recognition models can occasionally hallucinate words that were never spoken, especially during silence or noise, a known issue with some Whisper deployments. Classic recognition systems made different errors, such as misheard words. When accuracy matters, as in medical or legal transcripts, review outputs, use models with timestamps and confidence scores, and keep original audio for verification.
Does speech to text count as AI?
Yes. Speech recognition has been an AI research area for decades. Early systems used statistical models such as hidden Markov models, and modern systems use deep neural networks trained on vast amounts of audio, making them clearly AI.
The field progressed from recognising single digits in the 1950s to large-vocabulary dictation in the 1990s, then deep learning breakthroughs in the 2010s that brought accuracy close to human transcribers on clean audio. Today's voice assistants, live captions and dictation features all rely on neural speech recognition. Whether a product markets itself as AI or not, if it converts natural speech to text, it uses machine learning.
Is text to speech on CapCut AI?
Yes. CapCut's text-to-speech uses AI-generated neural voices that read your on-screen text or script aloud, with different voice styles and languages. It also offers AI voice effects and, in some versions, voice features that mimic speakers.
Add text, select it, and choose text to speech to generate narration on the timeline. Voices vary in quality and language coverage, and available voices can differ by region and app version. Check CapCut's terms for commercial use, especially for advertising or client videos. Avoid using voice features to imitate real people without consent, which can breach platform rules and laws. Many creators record their own voice for authenticity.
Can you use speech to text on a MacBook?
Yes. Turn on Dictation in System Settings, Keyboard, then press the microphone key or the Globe key twice in any text field to start speaking. On Apple silicon Macs, dictation in many languages runs on the device.
Also asked as: can i speech to text on macbook
You can speak punctuation and commands such as new line, and edit text with the keyboard while dictating. For full hands-free control, Voice Control lets you operate the Mac and dictate with voice commands. Accuracy improves with a good microphone and quiet environment. For long recordings, such as lectures or interviews, transcription tools or apps with Whisper models produce transcripts from audio files, with timestamps for review.
Can you use speech to text in Outlook?
Yes. Outlook for Microsoft 365 includes Dictate, available in new Outlook for Windows, Outlook on the web and Mac versions, letting you speak emails instead of typing. On Windows, voice typing also works with Windows key plus H.
Also asked as: can i speech to.text in outlook
Start Dictate from the message toolbar, speak naturally and say punctuation, such as comma or full stop. Dictation needs an internet connection and supports many languages. On phones, use the keyboard's microphone in the Outlook app. Outlook's Read Aloud does the reverse, reading emails to you. Pausing briefly between sentences often improves punctuation accuracy. Review dictated emails before sending, since names, numbers and technical terms are common error points.
Can you use speech to text in Notes?
Yes. On iPhone and iPad, tap the microphone key on the keyboard in the Notes app to dictate. Recent iOS versions also let you record audio in Notes with a live transcript. On Android, Google Keep supports voice notes with transcription.
Also asked as: can you speech to text in notes
On Mac, use Dictation in Notes like any text field. On Windows, Sticky Notes and OneNote accept voice typing with Windows key plus H, and OneNote also offers its own dictation and transcription. Voice notes are useful for capturing ideas while walking or after meetings, but review transcripts for errors. For organised meeting notes, transcription tools that separate speakers and produce summaries save more time.
What is text to speech used for?
Text-to-speech is used for accessibility, such as screen readers for blind and low-vision users, reading support for people with dyslexia, audiobooks and article narration, video voiceovers, navigation directions, language learning, customer service phone systems and voice assistants.
It also helps writers proofread by hearing their text, students review notes while commuting, and businesses produce training content in many languages. Podcasts and news apps increasingly offer narrated versions of articles. In AI agents, text-to-speech gives voice to conversational assistants. For accessibility, built-in operating system readers remain essential, while neural cloud voices suit polished narration. Disclose AI-generated voices in published content where audiences might assume a human speaker, and respect voice licensing terms.
What is text to speech tracking?
Text-to-speech tracking usually means highlighting each word or sentence as it is read aloud, so readers can follow along visually. Tools such as Microsoft's Immersive Reader, Speechify and many e-reader apps offer it.
Tracking helps people with dyslexia, language learners and young readers connect written and spoken words, improving comprehension and focus. Technically, apps use word timing information from the speech engine, such as boundary events in the browser's Web Speech API or timestamps returned by cloud voices, to highlight the current word. Adjustable speed and colour options help readers find a comfortable setup. Developers should test synchronisation across voices and speeds, since mismatched highlighting quickly becomes distracting.
Sources
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. For campus sessions the inviting institution covers the visit end to end: travel (a First AC train ticket for journeys up to five hours, a flight for anything longer), accommodation, meals, and pickup and drop; the honorarium and workshop fees are negotiable on WhatsApp or on the call, and requests are welcome from institutions that can cover the visit this way. Email pranjulrathour41@gmail.com.
