Logo
PhonerLite
VoIP softphone for Windows
📥

AI Settings

PhonerLite integrates AI in two practical ways, covering both conversational assistance and speech transcription:

Privacy and data flow

Before enabling these features, verify whether call content, recordings, and transcripts may be processed or stored in your environment.

Gemini

Gemini is a family of AI models developed by Google. They can understand and generate text, code, images and more. In order to use them, you need a Gemini API key. PhonerLite then accesses Google Gemini's online service via the REST API.

The required API key (access token) can be created free of charge here: https://aistudio.google.com/app/api-keys

In order to be able to refer to previous questions and answers from the dialogue, a certain history length should be set. A list of available models can be downloaded using the ⟳ button.

If a valid API key is stored and an Internet connection is available, this can be automatically integrated into a script using CLI (for example, using the "Ask" command). Alternatively, you can test this interactively directly on the "Test" page.

Ollama

Ollama is an open-source tool that simplifies the local execution of large language models (LLMs) such as Llama 3 or Mistral on your own computer. Since the models run directly on the local device after a one-time download, the AI can be used offline, ensuring maximum data control and privacy.

An Ollama server is typically installed by running a simple installation script or installer, which sets up a local service. By default, this server is accessible at the REST API address http://localhost:11434, allowing applications to interact with the downloaded local language models. The necessary hardware depends directly on the size of the language model (LLM) to be run locally with Ollama. For common models (such as 7B parameter models), a minimum requirement of 16 GB RAM and preferably a dedicated graphics card (GPU) with at least 8 GB VRAM is recommended, as the GPU greatly accelerates processing speed.

Transcribe

Transcription refers to the process of converting spoken language into written text. This is done automatically using speech recognition software (speech-to-text). PhonerLite uses Whisper for this purpose, which can be downloaded via the "Install" button (465 MB).

Additional models ("Medium [1.42 GB]" and "Large [2.88 GB]") can be installed manually. Although these require significantly more CPU or GPU resources, they offer a lower word error rate.

TTS

TTS (Text To Speech) refers to the conversion of text into spoken language. PhonerLite supports two types of TTS:

When Piper is downloaded, multiple voice models are installed depending on the package.

Important licensing note: Individual voices can have their own dataset licenses. These may impose stricter usage limits than the Piper software license itself.

After download, some voices are available and can be tested using the Play button.

Voices from integrated Windows TTS start with "Microsoft", downloaded voices start with "Piper".

Give It a Try

On this page, you can interactively engage in a dialogue with an AI counterpart (Gemini or Ollama). To test the transcription, press and hold the ▶ button while holding down the Ctrl key. During this time, the microphone signal will be converted into text.

AI Chat