Open source · No coding required · Built by researchers

Put your participants on a call with AI.

VIVA Studio lets you run studies where participants talk to an AI — set up as easily as an online survey. Pick a call interface, choose a voice, define your conditions, and share a link. Point-and-click all the way: no coding, and nothing to install for you or your participants.

No coding required Free & open source Your data stays yours
Random assignment Condition B · pitch −2
Transcript & recording secured
Prolific ID captured
Works with cutting-edge AI models
OpenAI Google Gemini Anthropic ElevenLabs + any custom API
What you get

Everything you need. Nothing you need to code.

VIVA Studio is built for social scientists — marketing, management, psychology — who want to study how people talk to AI without becoming programmers along the way. If you can set up an online survey, you can run a VIVA Studio study.

Set up a study in minutes

Four guided steps, just like building a survey — from consent page to shareable participant link in one sitting. No code, no installs, no ticket to your IT department.

Cutting-edge AI models

Use the latest voice models from OpenAI, Google, Anthropic or ElevenLabs — and switch between them with a click, so you can compare providers instead of marrying one.

Make the voice your manipulation

Raise or lower the pitch, speed it up or slow it down — the same AI can sound like a different persona in every condition.

Conditions & randomization built in

Define as many experimental conditions as your design needs — participants are assigned automatically, or forced via the study link.

Fits into your survey flow

Send participants over from Qualtrics or Prolific and back again — IDs and any other variables travel along automatically. See how ↓

Interfaces, ready to use

Pick a pre-designed call screen and adjust colors, text and layout in a point-and-click editor. See the interfaces ↓

An AI assistant for the design

Describe the look you want in plain words and the built-in assistant restyles the interface for you — no design skills needed.

Data ready for analysis

Transcripts, recordings, call durations and URL variables — one click away, formatted for R, SPSS, Stata or Python.

Collected data is encrypted — only you can open it

Participants' transcripts and recordings are sealed on their device and unlocked with your personal key — end-to-end with most model setups. During the conversation itself, speech is processed by the AI provider you selected. More on security ↓

How it works

From idea to live study in four steps.

The study manager walks you through the same four steps for every project — each one previews exactly what your participants will see. If you can build a survey, you can do this.

1

Set up the study

Name your study, define experimental conditions, and write the welcome, consent and end-of-call pages with a familiar rich-text editor.

ConditionsConsentRandom assignment
2

Design the call interface

Pick a template, adjust instructions, timers and post-call questions — or restyle it with the visual editor and AI assistant.

TemplatesVisual editorLive preview
3

Choose the AI

Select a voice model, give it instructions per condition, pick its voice — and shift pitch or speed if the voice itself is your treatment.

Any providerVoicesPitch & speed
4

Go live & collect

Share the participant link, pipe in IDs from your panel or survey tool, and watch transcripts and recordings arrive — securely stored.

Study linksURL variablesCSV · JSON · ZIP
Qualtrics & co.

Lives inside your survey — or on its own.

Two ways to get participants on the call. Both take minutes to set up — pick what fits your study.

Or standalone, via link

Prefer it separate? Share the study link on its own — through Prolific, e-mail or a QR code — and send participants back with a completion code when the call ends.

The interfaces

Call screens your participants already know.

Every study starts from a ready-made interface — pick the framing that fits your research question, then make it yours. These are live previews of the built-in templates, just without the calling part.

Classic phone

A familiar smartphone call screen — the safe default when your study frames the AI as a phone call.

Live transcript

The spoken conversation appears as readable text in real time — ideal when reading along matters.

Video call

A meeting-style window with an avatar tile and speaking indicator — no camera involved, for you or them.

Every template is a starting point. Change colors, text and layout in the visual editor, or ask the AI design assistant — and instructions, timers and post-call questions come along.

Security & ethics

Participant data only you can read.

Voice data is sensitive. VIVA Studio treats it that way by default — encryption is always on, not an optional checkbox.

  • Your results are locked to you — transcripts and recordings are sealed on the participant's device and only your personal key opens them (end-to-end with most model setups).
  • A recovery key from day one — lose your password, not your data.
  • Your AI accounts stay safe — API keys are stored encrypted and never shown to participants; live calls use short-lived access tokens instead.
  • Consent, built in — editable consent pages and granular data-collection switches support GDPR-conscious, IRB-friendly designs.
Your participant speaksTheir transcript & recording are locked on their own device, with your key.
Your server stores only locked dataNobody with server access can read it — not a host, not an admin.
You unlock & analyzeEnter your password, read the transcripts, export for R or SPSS.

During the call itself, speech is processed by the AI provider you selected — choose the provider that fits your ethics protocol. (Under the hood: RSA-2048 + AES-256-GCM.)

The team

Built by researchers, for researchers.

VIVA Studio grew out of our own studies on how consumers talk to AI — and out of frustration with how much engineering stood between a good research idea and a running voice study.

Portrait of Marc Becker

Marc Becker

Portrait of Jana Holthöwer

Jana Holthöwer

Portrait of Daniel Wentzel

Daniel Wentzel

For your paper

Using VIVA Studio in your research?

We're glad it's useful! Please cite the platform so other researchers can find it — and so we can keep justifying the time we put into it. A methods paper is in preparation; the reference will be updated once it is published.

APA (7th ed.)

Becker, M., Holthöwer, J., & Wentzel, D. (2026). VIVA Studio: An open-source platform for conducting voice-AI research [Computer software]. https://github.com/voicelab

RIS imports directly into EndNote, Zotero, Mendeley and Citavi.

Questions

Frequently asked.

How much does VIVA Studio cost?
VIVA Studio itself is free and open source. You pay the AI providers directly for what your participants use — with current pricing, a typical five-minute conversation costs a few cents. Your API keys, your bill, no middleman.
What do I need to host it?
Any ordinary webspace with PHP 8 and MySQL — the same stack most university and shared hosts already provide. Upload the files, run the installer, done. An optional Node.js relay further reduces audio latency if your host supports it.
Which AI models can I use?
Integrated realtime models (OpenAI Realtime, Google Gemini Live) for the most natural conversations — or build your own pipeline from OpenAI or Google speech-to-text, OpenAI, Anthropic or Google language models, and OpenAI, ElevenLabs or Google voices. A custom connector lets you attach any other HTTP API.
Is participant data really safe?
Collected transcripts and recordings are encrypted: sealed in the participant's browser with your public key and decrypted only in yours — end-to-end with most model setups. The server, and anyone who might access it, only ever sees ciphertext. During the conversation itself, audio is processed by the AI provider you choose; nothing else leaves your infrastructure.
Does it work with Prolific, Qualtrics & co.?
Yes. Pass participant IDs and any other variables through the study URL — they're stored with each session and included in your exports. Conditions can be assigned randomly or forced via URL, so VIVA Studio drops neatly into your existing survey flow.

Ready to put AI on the line?

Download VIVA Studio, upload it to your server, and run your first voice-AI study today — your participants already know how to answer a phone call.