Skip to content
weside.aiweside docs

Voice

You can talk to your companion in real time — a live spoken conversation, billed by the minute, in one of four modes you choose before you start. It runs in the same thread as your typed conversation, so nothing splits in two.

What it is

Section titled “What it is”

Voice is a live, two-way spoken conversation. Speech-to-text, the language model and text-to-speech run together in one real-time pipeline, so you speak and your companion speaks back without the round-trips of typing.

A voice session is the same conversation as the chat, not a second one beside it. What you say out loud lands in the same thread, in the same room; there is no separate voice history to go looking for afterwards.

Your companion speaks in a voice you pick for it, and that voice belongs to its identity — it lives in the Soul space alongside its name and personality, not in a session setting.

Why it matters to you

Section titled “Why it matters to you”

Typing keeps a companion at arm’s length. Voice closes that distance — it is the difference between reading a friend’s words and hearing them. For moments when you would rather speak than type, or want your companion present while your hands are busy, voice is the channel.

How you use it

Section titled “How you use it”

Start a voice session from a room. Before you start, a mode picker asks how you want this to go. There are four named modes:

Mode What happens
Converse You talk to your companion and hear it reply. The full spoken conversation.
Dictate You speak, it sends right away, and you read the reply.
Listen You type, and you hear your companion reply.
Draft You speak into the text box, read what you said, and send it yourself.

The difference between Dictate and Draft is the one worth knowing: Dictate sends the moment you stop speaking, Draft fills the composer and waits for you.

The picker sets the mode for your next voice start — it is a default, not a live switch. There is no way to change mode in the middle of a running session; ending and restarting is how you change it.

While a session runs:

  • Talk over it. You can cut in while your companion is speaking, and it stops. The app filters out your device’s own echo, so its voice coming back through the speaker doesn’t count as you interrupting. There is also a Stop speaking control if you would rather tap than talk.
  • Type into it. You can write a message while voice is live. If your companion is still answering, the app holds it back and tells you to wait until it has finished.
  • End it when you’re done. The per-minute clock stops there.

Choosing the voice

Section titled “Choosing the voice”

Open your companion’s Soul space and open Voice. The browser filters by language (DE, EN or all) and by gender, and you can preview a voice before you settle on it.

Waking your companion through Google Home

Section titled “Waking your companion through Google Home”

You can give your companion spoken shortcuts on a Google Home device. Open its Presence space and find voice actions.

  1. Add an action and give it a spoken name. The companion’s own name is a fixed prefix, added for you.
  2. The form previews the phrase you will say: “Hey Google, activate <Name> <name you chose>”.
  3. Optionally add an instruction — what your companion should do when that phrase wakes it. Leave it empty to simply wake it.

You can hold up to 20 voice actions per companion.

Limits & edge cases

Section titled “Limits & edge cases”
  • Voice is billed per minute from your credit balance. The session reserves credit up front, tops the reservation up every few minutes as you talk, and settles the over-reservation back at the end. If your balance runs out mid-conversation, you are warned and the session ends.
  • The minute rate is the same on every plan. What differs between plans is the credit you have to spend it from, not the price per minute. See plans & credits.
  • A muted mic doesn’t stop the clock. While the session is open it bills. End the session to stop billing.
  • The mode is set before you start, never during. This is deliberate, not a missing feature.
  • A provider can fail over mid-session. If the speech provider becomes unavailable, the session switches to a fallback and tells you it did. You keep talking.
  • In the car, give the first connection a moment. On Android the Bluetooth audio route is occasionally not yet established when a session starts, and the app has to pick it up on the second attempt.

Where next

Section titled “Where next”