how to make wakeword " hey hope"?

[hsoftxl]

i type hey_hope , when training my model.
but not work?
how to train my "hey hope" model?

thanks

[muhammadabidPro]

Hello @hsoftxl!

Thanks for reaching out. That's a great question, and I'd be happy to help you get your "hey hope" model working perfectly!

The issue you're facing is a very common one and is easy to fix. It's all about how the target_phrase is written in your configuration file.

---

💡 The Core Problem: How the System Reads Your target_phrase

The target_phrase in your YAML file isn't just a label; it's the exact text that our Text-to-Speech (TTS) engine speaks to create the audio for your wakeword.

• When you write hey_hope, the TTS engine tries to pronounce "hey underscore hope," which isn't a natural phrase and results in incorrect audio data.
• To fix this, you need to provide the phrase exactly as a human would say it: "hey hope".

---

✅ The Solution: One Configuration, One Command

You can complete the entire process—from data generation to model training—in a single run. There's no need to edit the file multiple times or run multiple commands.

Here’s how to set up your config.yaml to do everything at once:

# A good, descriptive name for your model files. Underscores are great here.
model_name: "hey_hope_v1"

# A dedicated folder for this model's assets.
output_dir: "./trained_models"

# This is the most important part!
# Use the natural phrase with a space. The system needs this to generate correct audio.
target_phrase:
  - "hey hope"

# --- Pipeline Control ---
# Set all three stages to 'true' to run the entire process from start to finish.
# The system is smart enough to run them in the correct order (Generate -> Transform -> Train).
generate_clips: true
transform_clips: true
train_model: true

# Set this to 'false' to avoid re-generating feature if you run it again.
# If you want to start completely fresh, you can set it to 'true'.
overwrite: false

Run the Training

With your config.yaml file saved with the settings above, just run this single command in your terminal:

nanowakeword-train -c path/to/your/config.yaml -T

(Note: The -T flag is technically optional if train_model: true is set in your config, but it's good practice to include it to show your main intent.)

This command will automatically:

1. Generate the audio clips for "hey hope".
2. Transform them into feature files (.npy).
3. Train your final model.

That’s all there is to it! This single-step process should get your "hey hope" model trained correctly.

Let me know if you have any more questions. Happy training! 🚀

[hsoftxl]

target_phrase:

• "hey hope"

i add this config.

when i say these words, it will be waked.
Hope is what keeps us moving forward.
希望让我们不断前行。

I hope my efforts will pay off someday.
我希望我的努力终有回报。

She hopes to become a doctor in the future.
她希望将来成为一名医生。

We all hope for a better tomorrow.
我们都期待更美好的明天。

that is not normal.

i just want to wake up by "hey hope" or "hi hope" etc..

@muhammadabidPro

[hsoftxl]

i tried dnn rnn crnn e_branchformer, they all have this problem.
@muhammadabidPro

[muhammadabidPro]

If you ever need it again... Train DNN or RNN architecture