Best Brazilian Portuguese TTS voice for a talking avatar: how to choose and what to consider
How to choose your talking avatar's voice in Brazilian Portuguese: a ready-made voice or your own cloned voice. Quality test, how to record, why a recording gets rejected and a real example.
The voice decides whether a talking avatar looks like a person or a robot. It also drives the mouth: the video model follows the audio to move the lips. This guide covers Brazilian Portuguese (pt-BR) voices. In AvataReel, you have two options: a ready-made voice from the catalog or your own voice, cloned from a short recording.
Listen to an example with a ready-made voice. Helena, from the catalog, speaks a 13-second script generated from text:
What TTS is and why it affects the talking avatar
TTS (Text-to-Speech) turns text into speech. In a talking avatar, the path looks like this:
- You write the script.
- TTS generates the audio with the voice you chose.
- The video model uses the photo and that audio to generate the mouth, the blinks and the head movements.
That's why quality starts with the voice. If the audio has odd pauses, the mouth will too. See the full process in AI lip sync: how a photo becomes a talking avatar.
The problem with non-native voices in Brazilian Portuguese
Many tools were born in English and adapted their voices to Portuguese. The most common errors:
- Stress on the wrong syllable: "vídeo", "fácil" and "público" come out with the stressed syllable swapped.
- English rhythm: pauses between words that Brazilian Portuguese links together.
- Questions with no intonation: the sentence doesn't rise at the end.
- Nasals and digraphs: "ão", "ãe", "lh" and "nh" come out wrong.
In an ad, these errors cost you credibility in the first seconds.
Quick quality test
Before you choose a tool or a voice, generate this text and listen:
"Olá! Hoje vou te mostrar como aumentar suas vendas com menos esforço. O produto chegou em dois dias, sem complicação. Não é incrível?"
(Hello! Today I'll show you how to increase your sales with less effort. The product arrived in two days, no hassle. Isn't it incredible?)
| Criterion | What to listen for |
|---|---|
| Stress | "incrível" sounds like "in-CRÍ-vel" |
| Rhythm | Words flow, with no artificial pauses |
| Question | "Não é incrível?" rises at the end |
| Nasals | "vendas" and "complicação" sound like Brazilian Portuguese |
| Emotion | The voice varies in intensity, it isn't flat |
The two options in AvataReel
| Ready-made voice | My voice | |
|---|---|---|
| What it is | Brazilian Portuguese voices from the catalog | A voice similar to yours, created from your recording |
| What you do | Listen to the samples and choose | Record or upload at least 10 seconds |
| When to use it | Starting fast, catalog avatars, a fictional person | The avatar is you, or the brand has a well-known voice-over |
| Watch out for | It isn't your voice | Quality depends on the recording |
When you create an avatar by describing a fictional person, AvataReel suggests the ready-made voice that matches the photo.
How to clone your voice: step by step
- In Avatars, tap Create with my photo and upload the photo.
- In the Voice section, choose My voice.
- Tap Record and read the Script to read that appears on screen. Or tap Upload audio and send an MP3, M4A, WAV, OGG or MP4 file, up to 25 MB.
- Record at least 10 seconds. The bar marks 45 seconds as enough and stops on its own at 90.
- Confirm the authorization, review the photo and create the avatar.
In about a minute, the recording gets checked. When everything is fine, you see Your voice is ready.
How to record so the voice sounds like you
- Record in a quiet place, with no TV, fan or street noise in the background.
- Hold your phone a hand's width from your mouth, with no Bluetooth headset.
- Speak the way you do in your videos: natural tone, unhurried and without long pauses.
- Read 2 or 3 scripts in a row. We use the cleanest parts of the recording.
If "Too loud. Move the phone a little away from your mouth." appears while you record, move the phone away before you continue.
Why a recording can be rejected
If the recording doesn't come out well, you see Record your voice again with the reason. In the meantime, the avatar already creates videos with a voice similar to yours.
| Reason | How to fix it |
|---|---|
| Background noise | Record in a quiet place, with the phone a hand's width from your mouth |
| Distorted audio | Move the phone a little away and record again |
| Muffled voice | Record straight into the phone's microphone, with no Bluetooth headset |
| Words hard to understand | Speak unhurried, with the phone close to your mouth |
| Long silences | Read the script in one go, without stopping between sentences |
| Short or choppy recording | Speak continuously for longer |
Your first avatar with your photo is free. Record 10 seconds and you're done.
Create my avatar →Script: punctuation the voice understands
The voice reads punctuation. Use it to your advantage:
- Comma: short pause. Use it to breathe between ideas.
- Period: longer pause. It separates blocks of argument.
- Ellipsis: suspense pause. Useful before an invitation or a reveal.
- Exclamation mark: raises the intonation. Use it sparingly.
- All caps: avoid. It can distort the intonation.
Brand names, acronyms and numbers deserve attention. Write them the way they're spoken when the pronunciation isn't obvious. See examples in Pronunciation of names and acronyms in TTS.
Example of a script with functional punctuation:
"Você já perdeu uma venda... só porque não tinha vídeo? Escreve o roteiro, escolhe a voz e o avatar fala por você. Testa hoje, o primeiro avatar é por nossa conta."
(Have you ever lost a sale... just because you had no video? Write the script, choose the voice and the avatar speaks for you. Try it today, the first avatar is on us.)
Does the voice change with video quality?
No. Premium and Economy use the same audio. Lip sync is also the same. What changes is the image detail, such as skin texture and the sharpness of the teeth. See the comparison in Premium or Economy.
Start with your first video
The best way to judge the voice is to hear your own script. Create an account, no card needed, and build your avatar. The first one is free. Videos use credits, starting at 4 credits in Economy.
For the full picture, read the talking avatar guide, see how to create a talking avatar from a photo and how to give voice direction for your brand.
Frequently asked questions
What is TTS and why does it matter for a talking avatar?
TTS (Text-to-Speech) turns text into speech. In a talking avatar, that audio drives the mouth: the video model follows the speech to generate the lip movements. A robotic voice makes the whole video feel artificial, even with good sync.
Can I clone my own voice for the avatar?
Yes. When you create the avatar, choose My voice and record or upload at least 10 seconds. The recording runs up to 90 seconds, and 45 is already enough. After that, the avatar speaks any script with a voice similar to yours.
Why was my recording rejected?
The most common reasons are background noise, audio so loud it distorts, sound muffled by a Bluetooth headset, long silences and speech that is hard to understand. AvataReel shows the reason and how to fix it.
Ready-made voice or my voice: which should I choose?
Use the ready-made voice to start fast or for catalog avatars. Use your own voice when the avatar is you and your audience already knows your voice.
Can I use someone else's voice?
Only with permission. When you create the avatar, you confirm that you have the right to use the image and voice you uploaded.
Upload a photo, record your voice and type the script. Your first avatar is free.
Create my free avatar →