Audio Features¶

iPassion AI Portal includes a suite of audio capabilities that allow users to interact with AI models using their voice, listen to AI responses spoken aloud, and even conduct real-time voice and video calls with AI. This page covers all audio-related features and their configuration options.

Admin — Audio Settings

Overview¶

Audio features in iPassion AI Portal are built around three core capabilities:

Speech-to-Text (STT) — Convert spoken audio into text for hands-free message input.
Text-to-Speech (TTS) — Have the AI read its responses aloud using a synthesized voice.
Voice and Video Calls — Engage in real-time, conversational interactions with AI models using audio or video.

These features are available directly within the chat interface and require no additional software installation. Configuration options for STT and TTS engines are managed by administrators through Admin > Settings > Audio.

Important: Speech-to-Text and Text-to-Speech features require a secure HTTPS connection in production deployments. These features rely on browser APIs and WebRTC capabilities that are restricted to secure origins by modern browsers. HTTP-only environments will not have access to microphone or audio playback features. Ensure your iPassion AI Portal instance is served over HTTPS before enabling audio features for your users.

Speech-to-Text (STT)¶

Speech-to-Text allows users to speak into their microphone and have their words automatically transcribed into the message input field. This provides a fast, hands-free way to compose messages without typing.

How to Use STT¶

Click the microphone icon in the message input area.
Grant microphone access if prompted by your browser.
Speak your message clearly.
The transcribed text appears in the message input field in real time (or after you stop speaking, depending on the engine).
Review and edit the transcribed text if needed, then send as normal.

Click the microphone icon again to stop recording at any time.

STT Engine Options¶

iPassion AI Portal supports multiple Speech-to-Text engines, configurable by an administrator:

Browser Default (Web Speech API)¶

Uses the speech recognition capability built into the user's browser.
No additional server-side configuration required.
Quality and language support varies by browser (best support in Google Chrome and Microsoft Edge).
Processes audio locally in the browser or via the browser vendor's cloud service.
Suitable for general use and quick deployment without backend dependencies.

Whisper (Local)¶

Uses OpenAI's Whisper model running locally on the iPassion AI Portal server.
Audio is processed entirely on-premises with no data leaving your infrastructure.
Provides high transcription accuracy across a wide range of languages and accents.
Requires sufficient server resources (CPU or GPU) to run the Whisper model.
Recommended for privacy-sensitive deployments or environments without reliable external internet access.

Whisper (Cloud / OpenAI API)¶

Uses the OpenAI Whisper API for transcription.
Audio is sent to OpenAI's servers for processing and transcription is returned.
Delivers fast, highly accurate results without requiring local compute resources.
Requires a valid OpenAI API key configured in Admin > Settings > Connections.
Best suited for teams that already use the OpenAI API and prioritize accuracy and speed.

Admin Configuration¶

Administrators can select and configure the STT engine from Admin > Settings > Audio:

Setting	Description
STT Engine	Choose between Browser Default, Whisper (Local), or Whisper (Cloud).
Whisper Model Size	(Local only) Select the Whisper model size: `tiny`, `base`, `small`, `medium`, or `large`. Larger models are more accurate but require more resources.
OpenAI API Key	(Cloud only) The API key used to authenticate with the OpenAI Whisper API.
STT Language	Optionally specify the language for transcription to improve accuracy (e.g., `en`, `ar`, `fr`).

Text-to-Speech (TTS)¶

Text-to-Speech allows iPassion AI Portal to read AI responses aloud. Users can listen to answers rather than reading them, which is useful for multitasking, accessibility needs, or simply a more natural conversational experience.

How to Use TTS¶

Speak a response: Click the speaker icon that appears on any AI message to have that message read aloud.
Auto-play: When using Voice Call mode, responses are played automatically without requiring a manual click.

Playback can be paused or stopped by clicking the speaker icon again while audio is playing.

TTS Engine Options¶

iPassion AI Portal supports several Text-to-Speech engines, each offering different voice quality, language coverage, and infrastructure requirements:

Browser Default (Web Speech Synthesis)¶

Uses the speech synthesis capabilities built into the user's browser.
Available without any server-side configuration.
Voice quality and available voices depend on the operating system and browser.
Supports many languages through OS-level voice packs.
Suitable for testing or deployments where voice quality is not a priority.

OpenAI TTS API¶

Uses the OpenAI Text-to-Speech API to generate high-quality, natural-sounding speech.
Supports multiple voices with distinct characteristics (e.g., alloy, echo, fable, onyx, nova, shimmer).
Requires a valid OpenAI API key.
Audio is generated on OpenAI's servers and streamed to the user.
Recommended for production use where voice quality and naturalness matter.

ElevenLabs¶

Integration with ElevenLabs' voice synthesis platform.
Offers a wide library of highly realistic voices and custom voice cloning options.
Requires an ElevenLabs API key configured in the admin settings.
Best for use cases where voice branding or custom voices are important.

Azure Cognitive Services (TTS)¶

Uses Microsoft Azure's Text-to-Speech service.
Provides a large selection of neural voices across many languages and locales.
Suitable for organizations already using the Microsoft Azure ecosystem.
Requires Azure API credentials configured in admin settings.

Per-Model Voice Selection¶

Administrators can assign a specific TTS voice to individual AI models in iPassion AI Portal. This means different models can have distinct voices, making multi-model conversations easier to follow by ear.

Configuration is available in Admin > Models > [Select Model] > Voice.
Select from any voice available in the configured TTS engine.
Users can also override the voice in their personal Settings > Interface > Audio preferences, if permitted by the administrator.

Admin Configuration¶

Administrators configure TTS from Admin > Settings > Audio:

Setting	Description
TTS Engine	Choose the engine: Browser Default, OpenAI, ElevenLabs, or Azure.
TTS Voice	Select the default voice for all AI responses.
TTS Model	(OpenAI only) Select the TTS model to use (e.g., `tts-1` or `tts-1-hd` for higher quality).
ElevenLabs API Key	API key for ElevenLabs authentication.
ElevenLabs Voice ID	The specific ElevenLabs voice identifier to use.
Azure API Key and Region	Credentials for Azure Cognitive Services TTS.
Playback Speed	Adjust the speech rate for TTS output.

Voice Calls¶

iPassion AI Portal includes a Voice Call feature that transforms the chat interface into a hands-free, real-time voice conversation with an AI model. Instead of typing and reading, you speak and listen — creating a natural, phone-call-like experience.

Starting a Voice Call¶

Open a chat with any AI model.
Click the phone icon in the chat toolbar (top-right area of the chat interface).
The interface switches to Voice Call mode.
Grant microphone permissions if prompted.
Begin speaking — your voice is transcribed and sent to the AI automatically.
The AI's response is generated and spoken aloud via TTS without requiring any manual action.

How Voice Calls Work¶

Automatic turn detection: iPassion AI Portal detects when you have finished speaking and automatically submits your message.
Hands-free operation: The full STT-to-AI-to-TTS pipeline runs without any button presses during the call.
Continuous conversation: The conversation history is maintained throughout the call, just like a regular chat session.
Transcript: A full text transcript of the conversation is saved and accessible after the call ends.

Ending a Voice Call¶

Click the hang-up button or press the phone icon again to end the voice call. The conversation is saved to your chat history.

Requirements¶

A working microphone connected to your device.
Browser microphone permissions granted.
iPassion AI Portal served over HTTPS in production environments.
STT and TTS engines configured and enabled by an administrator.

Video Calls¶

iPassion AI Portal supports Video Calls, enabling face-to-face video interaction with AI models that support vision capabilities. This feature combines live video input with real-time AI analysis and spoken responses.

Starting a Video Call¶

Open a chat with a vision-capable AI model (e.g., a model that supports image input).
Click the video camera icon in the chat toolbar.
Grant camera and microphone permissions if prompted.
The interface switches to Video Call mode, displaying your camera feed.
Speak naturally — the AI sees your video feed and hears your voice, responding with both text and spoken audio.

How Video Calls Work¶

Live video frames from your camera are captured and sent to the AI model as visual context.
Voice input is transcribed using the configured STT engine.
The AI model processes both the visual and audio input together to generate a contextually aware response.
TTS playback delivers the AI's response as spoken audio in real time.
This enables use cases such as visual assistance, real-time coaching, document review via camera, or interactive demonstrations.

Ending a Video Call¶

Click the hang-up button or the video camera icon again to end the session. The conversation transcript and key frames (if saved) are accessible in your chat history.

Requirements¶

A working camera and microphone connected to your device.
Browser camera and microphone permissions granted.
A vision-capable AI model selected (video calls are not available for text-only models).
iPassion AI Portal served over HTTPS in production environments.
STT and TTS engines configured and enabled by an administrator.

HTTPS Requirement for Production Deployments¶

Browser security policies restrict access to the microphone and camera APIs to secure origins only. This means:

Audio and video features work on https:// URLs and localhost (for local development).
These features will not function on plain http:// URLs in production.
Administrators must ensure the iPassion AI Portal instance is deployed behind a valid TLS/SSL certificate.

If users report that the microphone or camera icon is missing or greyed out, verify that the portal is being accessed over HTTPS. This is the most common cause of audio feature unavailability.

For self-hosted deployments, popular options for enabling HTTPS include: - A reverse proxy (e.g., Nginx, Caddy, Traefik) with a Let's Encrypt certificate. - A cloud load balancer with TLS termination. - A wildcard or domain-specific certificate from your organization's certificate authority.

Summary of Admin Audio Settings¶

All audio features are configured centrally by an administrator. The table below summarizes the key settings location and scope:

Feature	Admin Setting Location	Key Options
STT Engine	Admin > Settings > Audio	Browser Default, Whisper Local, Whisper Cloud
STT Language	Admin > Settings > Audio	Language code (e.g., `en`, `ar`)
TTS Engine	Admin > Settings > Audio	Browser Default, OpenAI, ElevenLabs, Azure
TTS Voice (default)	Admin > Settings > Audio	Voice name or ID from selected engine
TTS Voice (per model)	Admin > Models > [Model] > Voice	Any available voice from configured engine
Voice Call availability	Admin > Settings > Audio	Enable/disable for all users
Video Call availability	Admin > Settings > Audio	Enable/disable for all users

Users can adjust their personal audio preferences (such as playback speed or preferred voice, if permitted) from Settings > Interface > Audio.

Reference

For more technical details, see the Open WebUI — Speech-to-Text & Text-to-Speech documentation.