Speech Add-ins

Speech add-ins are optional packages that give existing controls a voice. They are ordinary tools and actions — you attach them to a ChatView or a TextEntry the same way you attach any other one — so nothing about the host control changes, and dropping the package removes the capability cleanly.

They ship separately because speech pulls in Shiny.Speech and its platform permissions. Apps that never dictate anything shouldn’t carry that weight.

Platform coverage

Shiny.Maui.Controls.SpeechAddins targets iOS, Android, Mac Catalyst, macOS (AppKit) and Windows. There is no plain net10.0 target, so the GTK/Linux head cannot reference it — Shiny.Speech’s net10.0 assembly only implements the browser engine, so there would be nothing behind it there.

• 
• 

Frameworks

.NET MAUI

Blazor

What’s in the box

Tool	Host	What it does
SpeechToTextTool	ChatView input bar (MAUI)	Listens via ISpeechToTextService and backfills the entry, optionally submitting when recognition completes
TextToSpeechBubbleTool	ChatView message bubble (MAUI)	Reads a message’s Body aloud via ITextToSpeechService
TextEntrySpeechToTextTool	Any TextEntry (MAUI)	Appends recognized speech to the field, with a listening indicator
TextEntrySpeechToTextButton	Any input (Blazor)	The same voice-input button, driven by the browser’s Web Speech API
PromptTextToSpeechTool	Quick Entry PromptView (MAUI and Blazor)	Reads the prompt’s answer aloud via ITextToSpeechService

Setup

.NET MAUI

Install Shiny.Maui.Controls.SpeechAddins and register Shiny Speech — AddSpeechServices() (or just AddTextToSpeech() for read-aloud on its own). Every tool here resolves its engine from DI and no-ops when nothing is registered, so a forgotten registration looks like a button that does nothing. The tools are registered under the same http://shiny.net/maui/controls XAML namespace as the core controls, so there is no extra xmlns to add.

Speech recognition needs a microphone permission entry in the app manifest — a library cannot add it for you. See Shiny.Speech for the per-platform requirements.

Blazor

Install Shiny.Blazor.Controls.SpeechAddins and add the @using — typically in _Imports.razor:

@using Shiny.Blazor.Controls.SpeechAddins

Recognition uses the browser’s Web Speech API, which is not implemented everywhere. Chromium-based browsers support it; Firefox does not. The button is a no-op where the API is missing rather than an error.

ChatView

Add the tools like any other action:

<shiny:ChatView Provider="{Binding Provider}" SessionId="{Binding SessionId}">
    <shiny:ChatView.InputActions>
        <shiny:SpeechToTextTool AutoSend="False" SilenceTimeout="00:00:03" />
    </shiny:ChatView.InputActions>
    <shiny:ChatView.CustomBubbleActions>
        <shiny:TextToSpeechBubbleTool />
    </shiny:ChatView.CustomBubbleActions>
</shiny:ChatView>

Or in code:

using Shiny.Maui.Controls.SpeechAddins.Chat;

this.InputActions = [ new SpeechToTextTool { AutoSend = false, SilenceTimeout = TimeSpan.FromSeconds(3) } ];
this.CustomBubbleActions = [ new TextToSpeechBubbleTool() ];

SpeechToTextTool	Default
AutoSend	false	submit the entry after recognition completes
SilenceTimeout	2s	silence before recognition is considered done
Culture	null	recognition culture (e.g. "en-US")
PreferOnDevice	false	prefer on-device recognition
ListeningText	⏹	glyph shown while listening

TextToSpeechBubbleTool	Default
SpeechRate / Pitch / Volume	1.0
Culture / VoiceName	null (system default)

TextEntry

TextEntrySpeechToTextTool is a TextEntryTool, so it sits in the entry’s tool strip beside things like ClearButtonTool:

<shiny:TextEntry Placeholder="Say something">
    <shiny:TextEntry.RightTools>
        <shiny:TextEntrySpeechToTextTool Culture="en-US" SilenceTimeout="00:00:02" />
    </shiny:TextEntry.RightTools>
</shiny:TextEntry>

TextEntrySpeechToTextTool	Default
SilenceTimeout	2s	silence before recognition is considered done
Culture	null	recognition culture
PreferOnDevice	false	prefer on-device recognition
ListeningText	⏹	glyph shown while listening
ListeningColor	—	tint applied while listening

The tool restores the entry’s original text if recognition is cancelled, so an aborted dictation doesn’t leave a half-transcribed field behind.

Blazor

<TextEntrySpeechToTextButton @bind-Text="message" Culture="en-US" />

Parameter	Default
Text / TextChanged	— (bindable)
IdleText / ListeningText	🎙 / ⏹
IdleColor / ListeningColor	#4CAF50 / #F44336
Culture	null
Continuous	false

Quick Entry

PromptTextToSpeechTool is a PromptTool, so it docks into a PromptView’s tool slots — the same shape the TextEntry tools above have. It ships on both hosts.

<qe:PromptView>
    <qe:PromptView.TrailingTools>
        <shiny:PromptTextToSpeechTool AutoSpeak="True" Culture="en-US" />
    </qe:PromptView.TrailingTools>
</qe:PromptView>

<PromptView Response="@answer" TrailingTools="tools" />

@code {
    readonly List<PromptTool> tools = new() { new PromptTextToSpeechTool() };
}

PromptTextToSpeechTool	Default
AutoSpeak	false	read the answer as soon as it lands
HideWhenEmpty	true	hide the tool until there is something to read
TextSelector	null	Func<PromptView, string?> — what to read
SpeechRate / Pitch / Volume	1.0	passed through to TextToSpeechOptions
Culture / VoiceName	null (system default)
SpeakingText / SpeakingColor	⏹ / #F44336	the stop-state glyph and tint (Blazor: SpeakingIcon)

It speaks PromptView.Response — the plain-text half of the answer. A rich answer set through ResponseContent has no text in it to hand a synthesiser, so give the tool a TextSelector that pulls the words out of what you rendered.

Blazor needs WebAssembly

Synthesis goes through Shiny.Speech’s browser implementation over speechSynthesis, so register AddTextToSpeech() and expect nothing on a Blazor Server host — the synthesiser is the visitor’s browser.

Next Steps

• Quick Entry — the prompt these dock into
• ChatView Custom Actions — how input actions and bubble actions fit together
• TextEntry — the tool strip these plug into
• Shiny.Speech — the underlying recognition and synthesis services