Generate Portuguese text to speech for Brazil or Portugal with guidance for nasal vowels, local vocabulary, numbers, names and per-project credit planning.
Portuguese text to speech needs one early decision: Brazil or Portugal. The shared language does not make every voice, word and rhythm interchangeable. A script that sounds natural in São Paulo may feel formal or foreign in Lisbon, and the reverse is equally true.
Choose by audience and publisher, not by population alone. Brazilian Portuguese is common in large consumer markets; European Portuguese can be essential for a Portugal-based brand or institution. Some projects use a neutral voice, but “neutral” does not guarantee that every listener hears their own usage.
Write down the choice before casting. Then check vocabulary, pronouns, verb forms, dates and pronunciation against that decision.
Differences appear in everyday words, forms of address and the rhythm of a sentence. A word that is ordinary in one country can sound dated in another. Translate for the market rather than converting spelling only.
Keep the local terminology in the final script and give the voice a complete sentence to work with.
Portuguese has nasal vowels and sound contrasts that can be difficult to judge from spelling alone. Generate a short paragraph containing the names, numbers and recurring technical words from the project. Listen for the contrasts that matter in context rather than trying to diagnose the entire sound system.
If a voice misreads one word, correct it in the script or add a test note for the reviewer. Do not assume the entire voice is unsuitable because one ambiguous name needs a clarification.
One credit covers about 1,000 characters. Natural speech rates vary, so build the estimate from the actual script rather than assuming that every five-minute video has the same word count.
| Asset | Words | Characters | Approximate credits |
|---|---|---|---|
| 30-second ad | 65–85 | 350–500 | 1 |
| 5-minute explainer | 650–800 | 4,000–5,500 | 5–6 |
| 20-minute lesson | 2,600–3,300 | 14,000–17,000 | 14–17 |
| 40 episodes | 30,000 | 160,000–180,000 | 160–180 |
The $9 Starter pack covers a first short project. A monetized channel, paid course or client deliverable requires the $29 Creator pack for commercial rights.
Pick one voice for the series and keep a pronunciation list for names, products and recurring phrases. The voice should sound like it belongs to the same brand in a TikTok clip, a product demo and a support answer.
Use the same market convention across every file. Mixing Brazil and Portugal in one campaign is confusing unless the difference is intentional and clearly part of the content. AI voice for TikTok and AI voice for e-learning cover different formats with the same casting discipline.
For dubbing, preserve the source timing and review the localized script against the original performance. The AI dubbing guide covers that workflow. A translated sentence that reads well on a page may be too long for the original shot, so adapt the wording before generating the voice.
Test names at the beginning because a wrong pronunciation is immediately noticeable. For numbers, decide how the figure should be read before generating:
Write the spoken form when the model has room to choose. For example, an abbreviation can be understood visually but read differently by voice.
Ask a native speaker from the target market to listen without reading the script. They should understand the offer, identify the brand and be able to repeat the action. Then compare the register and vocabulary with the audience.
Do not send a Brazilian referee to approve European Portuguese without labeling the market, or the reverse. Variant taste is part of quality.
Keep the Portuguese text available, label the audio duration and avoid autoplay. A transcript is useful for courses and public information. For documents, the reader guide covers preparation; for multilingual versions, see AI dubbing.
Commercial usage and voice consent remain separate requirements. The voice cloning consent guide explains the permission record.
Yes. Choose a Portuguese voice that matches the target market, test Brazilian and European pronunciation where relevant, and proof-listen the first paragraph before generating the full project.
Use the variant the audience expects. Vocabulary, rhythm and some grammar differ. A Brazilian-sounding read can feel wrong in Portugal, and a European read can feel distant to a Brazilian audience.
Numbers vary by context and variant. Test prices, dates, phone numbers and ranges before a full generation, and write ambiguous figures in the spoken form you want.
One credit covers about 1,000 characters. A five-minute video at roughly 750 words is around 4,500–5,500 characters, or 5–6 credits before alternate takes.
Commercial rights start with the $29 Creator pack. Personal projects are covered by Starter, while cloned voices require documented speaker consent.