How can I call tts.speak service with tts.piper?

Configuration Voice Assistant
1

The error I got

image
image1307×456 22.5 KB

I also tried en-us and en-us-amy-low. None of them work.

2

try this :point_down:

service: tts.speak
data:
  media_player_entity_id: media_player.your_media_player_with_a_very_long_name_come_here
  message: I am so cute
target:
  entity_id: tts.piper

:speaking_head: :speech_balloon:

3

This works for English. How can I use another language?

4

Set your voice and language in the Piper addon configuration.

image
image1155×451 17.8 KB

6

It doesn’t work. It still uses English.

7

Works fine here. Did you click save and restart the addon after changing the language?

Did you use a different language for the message parameter?

eg:
message: "Bonjour! Comment ça va ?"

8

Then choose a existing language you would like to use :point_down:

service: tts.speak
data:
  media_player_entity_id: media_player.your_media_player_with_a_very_long_name_come_here
  language: whateverexists
  message: I am so cute
target:
  entity_id: tts.piper
9

Yes and Yes. Have you updated to the latest version of HA? 2023.10.0

10

I don’t know which language code I should use for each language. I have updated my original post to show the error.

11

You also set the voice - check the piper add-on for the exact codes like shown in the screenshot from @jazzyisj

12

Yes. Piper is working fine, but I’ve also never specified the language in the service call. I just set the voice in the addon configuration. I did try to set the language to both “en” and “english” using a dev tools service call and got errors.

Have you tried just not setting the language parameter?

13

That’s why I posted this post, I’ve already checked this and tried en, en-us, and en-us-amy-low.
image

14

How would I use that in the basic Hass configuration, so I can use it for notifiers, etc.

I now have this in there:

# Text to speech
tts:
  - platform: google_translate
    language: "nl"
1 Like
15

Not sure if this is helpful or not but I found this thread while trying to figure out how to use tts.speak with Piper in an automation. I also select from multiple languages at random in a single automation.

Note that I am not passing a language value to the service. Selecting the voice (which includes the language already) is enough for it to work without selecting a language. In my example I use two voices:

  • zh_CN-huayan-medium | Chinese (Simplified) medium quality voice
  • en_GB-alan-low | British English low quality voice

The tts message is included in an array with the voice code and then one of these arrays is selected at random and played.

alias: Brush Teeth announcement
  description: ""
  variables:
    message: "{{ [
      ['请注意。你应该去刷牙。', 'zh_CN-huayan-medium'],
      ['男孩们，你需要去刷牙。马上！', 'zh_CN-huayan-medium'],
      ['10 位牙医中有 10 位建议刷牙和使用牙线。现在是你刷牙的时候了！', 'zh_CN-huayan-medium'],
      ['Boys, it is time to brush teeth. Please get to it.', 'en_GB-alan-low' ],
      ['It is time to brush your teeth. You dont want to have bad breath, do you?', 'en_GB-alan-low' ],
      ['Boys, please stop what you are doing and go brush your teeth now. Thank you.', 'en_GB-alan-low' ]
      ]
      | random }}"
  trigger:
    - platform: time
      at: "19:05:00"
  action:
    - service: tts.speak
      entity_id: tts.piper
      data:
        media_player_entity_id: media_player.den
        message: "{{ message[0] }}"
        options:
          voice: "{{ message[1] }}"
  mode: single
4 Likes
16

thank you mate.
here is an easier version: working in 2025


action: tts.speak
metadata: {}
data:
  cache: true
  media_player_entity_id: media_player.vlc_telnet
  message: >-
    The minimum temperature for the Marstek Battery on the Balkony is
    currently     {{
    states('sensor.hame_energy_hmj_2_7cfefewwef2f_temperature_min') }}    
    degrees, which is below 10 degrees. - spoken with piper
  options:
    voice: en-gb-southern_english_female-low
target:
  entity_id: tts.piper
17

piper as an addon and wyoming protocol in integrations. You can find a list of voice at rhasspy/piper-voices · Hugging Face

here is an example in an automation to talk to two media_players at once.


 - parallel:
            - service: tts.speak
              target:
                entity_id: tts.piper
              data:
                media_player_entity_id: media_player.squeezelite_garden_2
                message: "{{ alert_message }}"
                cache: true
                options:
                  voice: en_US-lessac-medium

            - service: tts.speak
              target:
                entity_id: tts.piper
              data:
                media_player_entity_id: media_player.dublin
                message: "{{ alert_message }}"
                cache: true
                options:
                  voice: en_US-lessac-medium

I also taught chatgpt so you can ask there

do you remember the conversation we had about piper

ChatGPT said:

Yes, I remember our previous conversation about Piper in Home Assistant. Here’s a quick recap of the key points we’ve covered:

:brain: What I Remember:

  • You’re using Piper with the Wyoming Protocol integration in Home Assistant.
  • You wanted to change the voice used by Piper.
  • I initially mentioned doing this via Devices & Services → Integrations, but you correctly pointed out:

“The default voice can be changed in Settings → Add-ons → Piper.”

  • You also confirmed that:

“Piper voices can be selected using the options.voice key, with values like en_US-joe-medium from the Hugging Face repo.”

  • You were using Whisper, OpenWakeWord, and Piper together.