Want to suggest a wake word? Leave your thoughts here. (AIS-1441)

Hi all,

We're excited to offer the community more free and high-quality wake word models. Everyone has their own unique wake word preferences. Now, we're ready to regularly release some of the most popular wake words. Please let us know the wake words you want!

Already supported languages: Chinese, English, Japanese, French, Spanish, Portuguese, German
Planned support for: Korean, Russian, and Arabic

In the past, it was an expensive process to collect high-quality human speech data. But now, our team has developed a cost-effective way to train wake word models by using only TTS samples, which reaches 90-95% accuracy compared to models trained by human-recorded samples.

The wake word models and esp-sr have the same license and are free for commercial use. If you want a more accurate and exclusive wake word, please use our wake word customization service.

Currently, we support over 20 wake words. You can choose any one wake word to test. Starting from August 1, 2024, to get a new wake word, you'll need to meet one of these requirements:

If you've got an ongoing project, kindly attach the project link along with a brief overview when submitting your request.
Your wake word has been liked or upvoted by more than five people.

We have released Wake Word Training by TTS V2.0, which improves the TTS model and pipeline. Now it can reach 95-98% accuracy compared to models trained by human samples.

TTS Pipeline V3 now supports wake word training for Chinese, English, Japanese, and French. Planned support includes Korean, Spanish, Portuguese, German, Russian, and Arabic.

Note: Please carefully read the Wake Word Submission Agreement. All submissions are deemed to have agreed to this agreement by default.

The product names, logos, and brands associated with the wake words listed in this software are the property of their respective owners. They are shown here solely to provide examples of wake words for users to understand and test Espressif’s Speech Recognition Framework, and do not imply any affiliation with or endorsement by their owners. This software is not affiliated with, endorsed by, or in any way officially connected to any trademark owner. Before any commercial use, you must ensure that you are the lawful rights holder of the relevant wake words or have obtained proper authorization from the lawful rights holder.

At the same time, everyone is reminded to register and protect the trademark of your own wake word.

The Willow team and community would love "Hey Willow". It's our domain name because we've been waiting for this.

Thank you very much for offering this option, it's very exciting!

The Willow team and community would love "Hey Willow". It's our domain name because we've been waiting for this.

Thank you very much for offering this option, it's very exciting!

I'm glad you like this. Since "hey" and "hi" sound pretty similar, sometimes people might not really notice the difference. So, I was thinking, maybe we could support both "hey willow" and "hi willow" for waking up the device. That way, whether you say "hey willow" or "hi willow", it'll still work. Of course, when we release the wake word model, we'll call it like "wn9_heywillow". What do you think about that?

Good idea!

My only concern would be overall reduced accuracy (wake reliability vs false wake). We've noticed quite a bit of false wake with Alexa. From what I've read the automated TTS approach has 90-95% the accuracy of the models trained on human samples. I like "two word" wake words because they tend to improve accuracy, I suspect a 100% "Hey Willow" wake word could result in equivalent or even improved accuracy with the TTS approach vs even human sample trained Alexa?

Of course we could always test this, even starting with a pure "Hey Willow" model, a pure "Hi Willow" model, and a merged model.

Thanks again for offering this!

Your concern may indeed happen. We will generate two words and test which model performs better.

"hey/hi willow" model:
Model name: wn9_heywillow_tts
FAR(False Alarm Rate): 1 times / 8 hours
RAR(Right Alarm Rate): 88%

Test dataset description:
The FAR dataset: This dataset contains a total of 64 hours of audio data, which includes audio collected from the internet and audio recorded using esp32-korvo boards.
The RAR dataset: This dataset is generated by multiple commercial TTS APIs, with a total of approximately 500 samples. These data and models were not used in the training process. However, due to the differences between TTS samples and human samples, please exercise caution when referring to the test results.

Guys, what you are doing is really great. We have created a smart speaker called Homai based on the esp32-s3. We trained the model ourselves, but it is resource-intensive and not so easy to integrate into the pipeline. Could you please add support for our word Homai [ho'mai]? Thank you in advance!

@AigizK

Hi @AigizK ,
The syllable of Homai only has two. It is difficult to reduce the probability of false triggering for monosyllabic and disyllabic phrases. We recommend selecting a 3-5 syllable phrase as the wake word.

@sun-xiangyu

Hi @sun-xiangyu
We have already launched a project with this name, so we can't change it significantly. But can we use the variant "homa ai", where the sound 'A' is pronounced long?

We have already launched a project with this name, so we can't change it significantly. But can we use the variant "homa ai", where the sound 'A' is pronounced long?

I'm sorry that our TTS model cannot specify a syllable to extend its pronunciation at the moment. This means that we cannot generate a large number of accurate “homa ai” phrases.

Hi! Thank you for this awesome solution! We are developing a smart voice assistant called Sophia. Would it be possible to have the wake word "Hi Sophia"? This would help our user experience drastically. Thank you in advance!

@PrathamG

Hi @PrathamG , I'm glad you like it. "Sophia" sounds like a wake word that can be used directly. I mean, maybe we don't need an extra prefix "Hi". I suggest we start with just "Sophia". If the performance is not satisfactory, then we can train another one with "hi Sophia". What do you think?

Sure, that sounds like a good plan! We can use only "Sophia" and test the performance first. Thank you

If possible, I also wanted to request the wake word "Little Sophia". We are still unsure about which wake word to use, and having both options will help us determine this via user testing.

申请唤醒词：豆哥豆哥

唤醒词：豆哥豆哥（中文，拼音 dou ge dou ge，四音节重复）
使用场景：个人自用智能语音助手，固件基于 ESP-IDF v5.3 + esp-sr v1.9.x（AFE + WakeNet9）
硬件：正点原子 ATK-DNESP32S3（ESP32-S3 N16R8，16MB Flash / 8MB Octal PSRAM）
现状：目前使用 WN9_XIAOLONGXIAOLONG_TTS 可正常唤醒，希望通过 TTS 训练定制「豆哥豆哥」
需要的话我可以补充测试音频或其他信息

谢谢！

Custom wake word request — 豆包豆包 (dòu bāo dòu bāo), personal non-commercial

Hello ESP-SR team,

I would like to request a Mandarin wake word model through the community TTS pipeline.

Wake word: 豆包豆包
Pinyin (per syllable): dòu bāo dòu bāo (dou4 bao1 dou4 bao1)
Syllable count: 4
Structure: AABB repetition
Language: Mandarin Chinese
Suggested model name: wn9_doubaodoubao

Target hardware

Item	Value
Chip	ESP32-S3
Flash	16 MB
PSRAM	8 MB octal, 80 MHz
Audio input	single analog mic via ES8311 codec, 16 kHz / 16-bit mono
Framework	ESP-IDF v6.1
ESP-SR	v2.4.7
Model storage	dedicated 8 MB SPIFFS partition, `srmodels.bin` loaded at runtime
Requested model family	WakeNet9 (`wn9_*`)

Current status

The device is a small desktop companion robot and it is already running and validated on real hardware. The wake-word path is fully working end to end:

esp_srmodel_init()
-> assets partition / srmodels.bin
-> WakeNet9 model
-> AFE (AEC + NS)
-> wake-word callback

The model currently in use is wn9_nihaoxiaozhi_tts, packaged by the standard build pipeline. I only need to swap the wake-word model; the audio pipeline, AFE configuration and application logic stay unchanged. A model distributed in the standard srmodels.bin / model-partition format, compatible with ESP-SR v2.4.7, would drop straight in.

Use case

This is a personal, non-commercial hobby project — a desktop robot I am building for myself. It stays at home and is used only by me and my family. It is not sold, not offered as a service, not used for any commercial demonstration, not redistributed, and I am not monetizing it in any way. I am asking for the model purely for my own private use.

Why this word

It is a four-syllable AABB word in common use, which is exactly the pattern the TTS pipeline handles best, and it is the word I actually want to talk to my robot with every day.

Acknowledgment

I understand that submitting a request here is deemed to accept the Wake Word Submission Agreement, and that the released model would carry the same license as ESP-SR. I am requesting personal use only and make no claim of ownership over the word.

Thank you!

Wake Word: NEXO
Pronunciation: NEK-so
Languages: Portuguese (Brazil) and English

Target hardware:

ESP32-S3
ESP32-P4

Preferred engine: WakeNet10
Current engine: WakeNet9
Current model: wn9_jarvis_tts

Training method: TTS Pipeline

@handsomewym

@handsomewym

双儿：wakenet10_tts4h12q8p1_双儿_3_0.59_0.608

Welcome to test and provide feedback on the test results.

@jstockdale

@jstockdale

Hey/Hi Muse: wakenet10_tts4h12q8p1_Hi/Hey Muse_3_0.6041_0.6148

i have buidl a new firmware for my circuitmess Butterbot. So i really would like to have "hey butterbot" as wakeword.

That would be great since "hi esp" works but does not really ´fit´.

Thanks in advance

Custom wake word request: “Hey Nimo” — Nimo / M5Stack StickS3

Hello Espressif ESP-SR team,

I would like to request an English TTS-trained WakeNet model for the wake phrase
“Hey Nimo”. Nimo is a small physical AI companion built on the M5Stack
StickS3. Wake-word recognition must run fully offline on the device.

Project details:

Project: Nimo physical AI companion
Repository: https://github.com/abiyarilangeroudi-sudo/Nimo (currently private;
access or additional source details can be provided if needed)
Target: ESP32-S3-PICO-1
Hardware: M5Stack StickS3, 8 MB flash, 8 MB PSRAM, ES8311 audio codec
Audio input: single microphone, 16 kHz, signed 16-bit mono PCM
Firmware: ESP-IDF 6.1, ESP-SR 2.4.7, Xiaozhi-derived firmware
Current working wake model: wn9_hiesp
Runtime: ESP-SR AFE/WakeNet
Model storage: dedicated 2 MB assets partition
Requested phrase: Hey Nimo
Language: English
Suggested model name: wn9_heynimo_tts3, or the appropriate compatible
WakeNet10 equivalent

The physical device already has a working end-to-end voice pipeline, an
independent device credential, and Nimo Device Protocol v1. We can test a
candidate model on real StickS3 hardware and report right-alarm rate, false
activations, detection distance, room-noise behavior, CPU/memory use, and model
size.

Could you please advise whether WakeNet9 or WakeNet10 is the appropriate
deliverable for this firmware and consider generating the model through the
community TTS pipeline?

I have read and agree to the Wake Word Submission Agreement referenced in this
issue. Thank you.

Wake word request: 你好小猫 (nǐ hǎo xiǎo māo) — Chinese (Mandarin)

Wake word: 你好小猫
Pinyin: ni3 hao3 xiao3 mao1 (ni hao xiao mao), 4 syllables
Language: Chinese (Mandarin)
Preferred model: WakeNet9 TTS — e.g. wn9_nihaoxiaomao_tts.
We are on esp-sr v2.3.1 (WakeNet9 only; a WakeNet10 model would require a component upgrade).
Target chip / board: ESP32-S3 N16R8 (16 MB flash / 8 MB PSRAM),
1.85" round LCD, dual-microphone array.
Project: EchoEar / 喵伴 — an open-source AI desktop companion built on the
Espressif College EchoEar reference design and xiaozhi-esp32 firmware.
Project link: https://oshwhub.com/esp-college/echoear
Current wake word: wn9_miaomiaotongxue_tts works reliably on this hardware;
we would like to evaluate 你好小猫 on the same device.
We are happy to test on real hardware and report FAR/RAR.

Thank you!

21. Is there any plans to upgrade 'Sophia' to WakeNet10 format for ESP32-S3? Could you please take this request if not already taken up?
Thanks!

any comments or update on this?

Thanks!

30. > 21. Is there any plans to upgrade 'Sophia' to WakeNet10 format for ESP32-S3? Could you please take this request if not already taken up?
> Thanks!
 any comments or update on this?
 Thanks!

Under training

Hi Espressif team,

First of all, thank you for the excellent work on esp-sr and for opening this thread to collect community wake word requests. We really appreciate the effort your team puts into making high-quality wake word models freely available.

About our project
We are Fujigo Soft, a software company based in Vietnam. We are currently building a voice AI product called Vivi Voice AI. You can find a short introduction here: https://fujigo-soft.com/2026/10/23/vivi-voice-ai-la-gi/

Our request
For the next build of our project, we would like to use the wake word "Hi vivi" (English pronunciation). We understand that esp-sr already supports English wake words, and we would like to kindly ask whether it is possible to:

Have "Hi vivi" trained and added to the supported wake word list, or

Get guidance on how to use your Wake Word Customization Service (or the TTS-based training pipeline) to obtain a model for "Hi vivi" that we can use in our product.

What we can provide

A brief project overview and the public link above.

Any additional documentation or clarification you may need about our use case.

Willingness to follow the Wake Word Submission Agreement and any licensing/trademark requirements you specify.

We fully understand the requirements mentioned in your post (project link + overview, or community upvotes), and we are happy to comply with whatever process is appropriate.

If the best path is the paid customization service, we would also appreciate information on pricing and turnaround time.

Thank you very much for your time and support. We look forward to hearing from you.

Best regards,
DuongMX
Fujigo Soft
https://fujigo-soft.com/

Wake-word request: 你好小小 — Mandarin, ni3 hao3 xiao3 xiao3.

I am developing a private personal voice-assistant prototype using M5Stack AtomS3R (ESP32-S3, 8 MB Flash and 8 MB PSRAM) with Atomic Voice Base. It currently runs WakeNet9 locally with wn9_nihaodongdong_tts2.

Could the community TTS pipeline provide a WakeNet9 model for “你好小小”? Please advise the minimum compatible ESP-SR version.

The project repository is private. I can provide a minimal technical description if needed. Thank you.

Hi, I would like to request a new wake word: "Hey Muse" (English).

Project overview: I am building a personal DIY voice gadget based on Meta's open-source Muse Gadget SDK (facebookincubator/muse-gadget-sdk), running on a Waveshare ESP32-S3-Touch-AMOLED-1.75C board (ESP32-S3, dual microphone, speaker, AMOLED display). The gadget pairs with the Muse app; it is currently push-to-talk only, and I want hands-free activation with "Hey Muse". This is a personal, non-commercial project. I don't have a public repo for my own build yet, so I am linking the upstream SDK repo as the project base: https://github.com/facebookincubator/muse-gadget-sdk

Request details:

Wake word: "Hey Muse" (English)
Target: ESP32-S3, WakeNet9 / WakeNet10, TTS-trained model is fine
Language: English

I understand the Wake Word Submission Agreement applies. This is for personal use only; I will not use it commercially. If a TTS-trained model is possible, I would be very grateful.

Thank you!

Hi Espressif team,

Wake word: Vesta
Language: Spanish (Latin America)
Target chip: ESP32-S3 (N16R8, with PSRAM)
Framework: ESP-IDF 5.5
Model family: WakeNet9 or WakeNet10, whichever you ship for Spanish on ESP32-S3

Pronunciation: two syllables, stress on the first, «BES-ta» (/ˈbes.ta/).
The V is the Spanish b sound, as in «beso».

Use case: a desktop voice assistant. The wake word is the product name.
It is said at the start or at the end of a short phrase.

Project: https://vestastudiovzla.com/
Overview: Vesta is a desk assistant for Venezuela. The device is an ESP32-S3
with a round display, a microphone and a speaker. The owner links it from
that website. The wake word is always the name of the assistant.

I have read and agree to the Wake Word Submission Agreement.
I am responsible for the legal use of this wake word.

Thank you!

"Hi Kal"

Navigation Menu

Repository navigation

Want to suggest a wake word? Leave your thoughts here. (AIS-1441) #88

541 remaining items

Custom wake word request — 豆包豆包 (dòu bāo dòu bāo), personal non-commercial

Custom wake word request: “Hey Nimo” — Nimo / M5Stack StickS3

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Want to suggest a wake word? Leave your thoughts here. (AIS-1441) #88

Description

Activity

kristiankielhofner commented on Dec 14, 2023

feizi commented on Dec 14, 2023

kristiankielhofner commented on Dec 14, 2023

feizi commented on Dec 14, 2023

feizi commented on Dec 28, 2023

AigizK commented on Dec 29, 2023

sun-xiangyu commented on Jan 3, 2024

AigizK commented on Jan 3, 2024

sun-xiangyu commented on Jan 4, 2024

PrathamG commented on Jan 9, 2024

sun-xiangyu commented on Jan 9, 2024

PrathamG commented on Jan 9, 2024

PrathamG commented on Jan 9, 2024

541 remaining items

zhanglanlazy commented on Oct 26, 2026

Jister24 commented on Oct 29, 2026

Custom wake word request — 豆包豆包 (dòu bāo dòu bāo), personal non-commercial

Tauser commented on Oct 29, 2026

sun-xiangyu commented on Oct 30, 2026

sun-xiangyu commented on Oct 30, 2026

BaDWidoWMakeR commented on Oct 30, 2026

abiyarilangeroudi-sudo commented on Nov 4, 2026

Custom wake word request: “Hey Nimo” — Nimo / M5Stack StickS3

wangyifei5132 commented on Nov 5, 2026

gparvind commented on Nov 5, 2026

sun-xiangyu commented on Nov 5, 2026

Du0ngMX commented on Nov 5, 2026

longyige commented on Nov 5, 2026

dud4946 commented on Nov 5, 2026

gdelgadontiveros commented on Nov 6, 2026

theunstarted635-web commented on Nov 7, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions