Voice & Avatars
MeetLoyd offers two distinct ways for agents to speak:
- Live conversation voice — built into every agent, no setup, no extra cost. Plays from your browser's neural synthesis.
- Cloned & branded voice + video avatars — purchased from the Store as addon subscriptions for richer experiences.
Pick the right one for the job. Most teams only need the first.
Live Conversation Voice (free)
Every agent — including Loyd — can hold a real-time spoken conversation in your browser, with no addon required.
How it works
Speech recognition and synthesis run on your device, using the neural voices already shipped with your browser:
| Browser | Voice technology |
|---|---|
| Chrome (any OS) | Google WaveNet |
| Edge | Microsoft Edge Neural |
| Safari (Mac / iOS) | Apple Siri voices |
No audio leaves your machine. The agent only ever sees the text transcript. This is the lowest-latency, lowest-cost path — and it's the default for every browser surface.
Picking a voice
Tenant defaults
A tenant owner can set the default voice for the whole organisation:
- Loyd Assistant > Voice in the dashboard
- Two pickers: Loyd's voice and Default voice for new agents
- Each lists every voice your browser has installed, grouped by language. Click Preview to hear it.
- Save — applies to everyone in the tenant who hasn't pinned their own.
Per-agent override
Individual agents can have their own distinctive voice:
- Agents > Edit an agent
- Scroll to the Speech voice section in the Review step
- Pick a voice or leave it set to "Use tenant default"
Per-device override
A user who prefers a specific voice on their personal laptop can pin it locally — that pin wins over both the agent and tenant defaults. Currently set via browser devtools (a UI toggle is on the roadmap).
Voice resolution order
When an agent speaks, the voice is picked in this order:
- The user's per-device pin (if any)
- The agent's own voice (if set)
- The tenant's default-agent voice (if set)
- A quality heuristic that prefers network-rendered voices and known good families
- The browser's built-in default
If a pinned voice isn't installed on a particular user's device, playback falls through to the next step automatically — no broken audio.
Browser support
The microphone button appears in chat surfaces where speech recognition + synthesis are both available. That's Chrome, Edge, and Safari on desktop and mobile. Firefox doesn't currently ship SpeechRecognition; the mic button is hidden there.
Cloned & Branded Voice + Avatars (addon)
For experiences that need a recognisable brand voice — for example a CEO clone for executive briefings, or a video avatar that attends meetings — MeetLoyd offers a separate addon-based pipeline.
What addons unlock
| Addon | What you get | When to use it |
|---|---|---|
| Voice profile | A cloned or designed voice rendered server-side, with consistent identity across calls | Brand voice; phone calls; voicemail; future video avatars |
| Avatar | An AI-generated digital persona that speaks via lip-synced video | Recorded video messages; future Meet/Teams attendance |
Both are purchased from the Store. Subscriptions are billed monthly or yearly.
Voice profile types
| Type | Description |
|---|---|
| Cloned | Created from an uploaded WAV sample via the voice onboarding flow |
| Designed | Generated or selected from MeetLoyd's built-in voice library |
Voice profiles are stored per tenant. Each agent can have at most one cloned voice profile assigned.
Avatar types
| Type | Description | Best for |
|---|---|---|
| Image-based | Created from a single image | Quick setup, consistent branding, lower cost |
| Video-based | Created from video footage | Most realistic results, natural expressions, premium experiences |
Video lifecycle
| Status | Description |
|---|---|
| Pending | Job queued, waiting to start |
| Processing | Video is being generated |
| Completed | Video is ready for playback |
| Failed | Generation failed |
Buying and assigning
- Go to the Store and browse the Avatars & Voices category
- Subscribe to a voice or avatar product
- From Agents > Edit, scroll to the Voice & Avatar section to assign it
Best practices
Live conversation works out of the box on Chrome, Edge, and Safari. Don't buy a voice addon unless you actually need a cloned or branded voice.
Pick a default Loyd voice and a default agent voice in Loyd Assistant > Voice. Individual agents inherit from the tenant — only override per-agent when an agent really needs a distinct voice.
Set the agent's language to match your users. The browser picks the best installed voice for that locale.
Shorter messages generate faster and cost less. Aim for under 30 seconds of speech. For real-time UX, display the text response immediately and load the avatar video asynchronously.
Pricing
- Live conversation voice — included in every plan, no extra charge. The synthesis runs on the user's device.
- Voice profile addon — billed monthly through the Store's addon subscription system.
- Avatar addon + video generation — addon subscription plus per-second billing on generated video. Pricing varies by plan tier; contact sales or check the Store for current rates.
Next: Learn about MCP Integration for connecting external tools and services.