Models
The Models page centralizes all AI models used by your agents across three tabs: LLM (language models), TTS (text-to-speech voices), and STT (speech-to-text providers).
LLM Tab
Built-in Models
Callem provides pre-configured LLM providers with multiple model options. These are available to all agents by default.
Custom Models (Bring Your Own Model)
Connect any OpenAI-compatible LLM endpoint to use your own models. This works with any provider exposing the standard chat completions API.
Adding a Custom Model
- Navigate to Build > Models > LLM
- Click Add Custom Model
- Fill in the form:
| Field | Description |
|---|
| Display Name | How the model appears in dropdowns (e.g. “My Llama 3.1 70B”) |
| Base URL | The API endpoint without /chat/completions (e.g. https://api.together.xyz/v1) |
| API Key | Your provider’s API key — stored encrypted server-side |
| Model ID | The model identifier (e.g. meta-llama/Llama-3.1-70B-Instruct) |
- Click Test Connection to verify the endpoint responds
- Click Add Model
Once added, the custom model appears in the LLM provider dropdown when configuring agents.
The Test Connection sends a minimal request (max_tokens: 5) to verify the URL, API key, and model ID are correct. Cost is negligible.
Managing Custom Models
Use the action menu (three dots) on each custom model to:
- Test Connection — verify the endpoint is still reachable
- Edit — update name, URL, or API key
- Duplicate — create a copy with “(copy)” suffix
- Delete — remove the model (agents using it will need reconfiguration)
TTS Tab — Voice Library
Built-in Voices
Browse all available TTS voices across providers. For each voice:
- Preview — click the play button to hear a sample
- Filter — by provider, gender (male/female), or search by name
Voice Cloning
Create custom voices from audio samples.
Cloning a Voice
- Click Clone Voice
- Fill in the form:
| Field | Required | Description |
|---|
| Reference Audio | Yes | WAV or MP3 file, 5-20 seconds, clean audio |
| Transcription | Yes | Exact text spoken in the audio sample |
| Voice Name | Yes | Display name (e.g. “Sophie - Commerciale”) |
| Language | Yes | Default language for the voice |
| Gender | Yes | Male or Female |
| Description | No | Internal notes (e.g. “warm tone, for solar campaign”) |
| Tags | No | Comma-separated labels for filtering |
- Click Upload & Clone
Cloned voices appear in the “My Voices” section and are also available in the agent Voice Browser when selecting a TTS voice.
Audio quality matters. Use a clean recording with minimal background noise, recorded at a consistent volume. Mono audio at 24kHz is ideal.
My Voices Filter
When you have cloned voices, a My Voices filter appears in both the Models page and the agent Voice Browser, letting you quickly find your custom voices.
STT Tab — Speech-to-Text
Browse available STT providers and their supported languages. Each provider card shows:
- Provider name
- Number of supported languages
- Expandable language list with flag, label, and code
Custom STT endpoints (Bring Your Own STT) are coming soon.
