Documentation›Models, images, and voice
34 Models and media

Voice Messages, Hands-Free, Stories, and Songs

Send a recorded voice message, start a call by saying her name, ask for a bedtime story, or ask for a song.

Local Waifu 1.8.0 macOS Windows
plain English first

Follow the bold button names and numbered steps. Technical names are included only when you need to recognize a model or file inside the app.

Chapter 33 covers dictation, spoken replies, and live calls. This chapter covers four more things she does with sound.

Send a voice message

Goal

Record a short message and send it as audio.

Before you start

Speech recognition must be installed, because the message is transcribed. Chapter 33 covers installing it.

Steps

  1. In a one-to-one chat, press and hold the voice message button next to the message box. The tooltip says Record a voice message.
  2. Talk while you hold it. The box shows Recording.
  3. Release to send. To cancel, slide up before releasing or press Escape.

Expected result

Your message appears as an audio bubble with a transcript, and she answers.

If something goes wrong

  • A press that is too short is not sent. Hold the button while you talk.
  • A message can be two minutes at most. If you hit the limit, the end is not recorded.
  • If a reply was still coming in, the message is not sent.
  • If the microphone disconnects, the message stops.
  • In a group chat voice messages do not work. Type instead.
  • If the transcript cannot be made, the bubble shows No transcript.

Start a call hands-free

Goal

Open a voice call by saying her name, without touching anything.

Steps

  1. Open Settings → Models → Voice Model and find Hands-free under The call.
  2. Turn it on.
  3. Check the Wake phrase. It is her name unless you change it. A phrase that is too short to tell apart from ordinary speech is refused, so use a longer one. Three letters are the minimum, or two characters for Chinese, Japanese, and Korean.
  4. Choose whether to Require “hey” before her name.
  5. If the tab offers it, download the small wake-word model.
  6. Say her name.

Expected result

The status shows Listening, and saying the phrase starts a call.

Privacy

The setting says it listens on your computer only: the audio is never saved and never leaves your computer, and the microphone is open only while it listens. It is off by default. The status reads Paused while the microphone is in use during a call.

If something goes wrong

If listening cannot start, the app shows Couldn’t listen with the reason and leaves the switch off.

Ask for a bedtime story

Goal

Have her tell a story out loud that stops by itself when you fall asleep.

Before you start

She tells the story in her voice, so install the voice model first. A story needs an active license and cannot start during a voice call.

Steps

  1. In the message area, open the story and song button and choose Bedtime story.
  2. Under What should it be about?, type a theme, or leave it empty.
  3. Choose a length: Short (about 3 minutes), Medium (about 8), or Long (about 15).
  4. Choose a Sleep timer or leave it off.
  5. Select Start the story.

Expected result

She tells the story in parts, and the panel shows Part 1 of N and the minutes left. She ends it with The story is over. Sleep well. Select Stop to end it early. You can set a default length and sleep timer under Settings → Models → Voice Model → Bedtime story.

If something goes wrong

  • End the voice call first, then start a story.
  • A story is already being told.
  • While a story plays, dictation and voice messages are disabled.
  • A story is not saved as memories beyond a single line.

Ask for a song

Goal

Have her write and sing a short song for you.

Before you start

Songs are made by a cloud service, so they need a fal.ai key. Add one under Settings → Models → Image Models. The same key is used for image generation. Songs also need an active license, and they do not work in Offline Mode.

Steps

  1. Open the story and song button and choose A song from her.
  2. Type a theme under What should it be about? and a Style, or leave them empty.
  3. Read the consent line: when you press Create, the lyrics, the style, and the theme are sent to fal.ai. Nothing is sent before that. Your chat model writes the lyrics first, so if it is a cloud model the theme and a few memories go there too.
  4. Select Create the song.

Expected result

She writes lyrics, then composes the song. It arrives as an audio bubble in the chat, with a Show lyrics toggle.

If something goes wrong

  • Songs need a fal.ai key. Select Open Image Models to add one.
  • Offline Mode is on. Turn it off first.
  • The music endpoint id in Settings is not valid. In Settings → Models → Voice Model → Songs, select Reset to default under Music endpoint.
  • You can also change Song length there.
  • Songs cannot start during a call, and only one song is made at a time.