Is Inworld Audio free?

How to make a short recording on Inworld's website

Start with one short original passage and a provider-designed preset. Our six examples show the browser route; you do not need to download model weights or set up a local speech server.

1. Check Billing first

   Read Free credits remaining, current plan, credit history and expiry. Keep auto-reload off if you only intend to evaluate Free access. A model announcement alone does not establish a funded request.

2. Match the voice and language

   Open Voices & Settings, then the voice selector. Use Inworld voices and its language filter. Choose a voice that explicitly supports your script's spoken language; our Flash interface warned against cross-lingual use. Check the language again after changing voices.

3. Choose the actual model

   Our selector offered TTS-2 and TTS-2 Flash. These recordings use inworld-tts-2-flash, with Standard quality, Delivery 1 and speed 1×. These are observed settings, not recommended settings for every text.

4. Generate once and review

   Paste the script, check language mismatch warnings and the applicable cost. Our editor did not show a per-request quote; the checked On-Demand Flash rate was $15 per million characters. Generate, compare the recording with its transcript and check Usage before deciding on any retry.

5. Download the original

   Choose MP3 in Download format, then use Download audio after completion. We saved six native MP3s. History remains local to the device/workspace, so a history entry alone is not a backup.

What our six-language test establishes

Sarah speaks English, Mateo Spanish, Hélène French, Haruto Japanese, Aanya Hindi and Murilo Portuguese. Each script describes a different small scene. The playable examples include the exact text so you can compare delivery without treating these short passages as a full language evaluation.

All six completed with TTS-2 Flash through the website. The Usage page recorded 844 characters at $15 per million characters, with a rounded $0.01 cost estimate. Billing displayed $1.00 before and $0.98 after; do not interpret that rounded difference as a precise charge. No paid credit purchase or auto-reload was used.

At the published rate, 844 characters is about $0.01266 before any different metering or rounding. That arithmetic is not an exact observed charge. This account test does not promise that every visitor receives the same grant, expiry or access.

Which current speech model did we test?

The current model list and our selector include Realtime TTS-2 and TTS-2 Flash. The release notes date TTS-2 to May 5, 2026 and Flash to August 9, 2026. Older 1/1.5 releases are legacy options rather than the target of these examples.

Flash is the model used in this gallery. In our interface it warned that a voice speaking outside its supported languages would sound accented and offered a switch to TTS-2. Select a matching preset instead of assuming every voice works equally well in every listed language. Provider latency and quality claims are not our measurements.

How far do the Free credits go?

Pricing advertises starter access with up to 70 TTS minutes, but that estimate uses Flash rates and approximate text length. It is not 70 guaranteed minutes on every model. The checked On-Demand rates were $25 per million characters for TTS-2 and $15 for Flash, with a 2,000-character Playground request limit.

Our account's $1.00 grant was labelled Free credits and had a visible expiry. We did not establish a recurring renewal, universal eligibility or rollover rule. Check your own history rather than applying the paid subscription credit rules to Free evaluation access.

Short scripts keep experiments inexpensive, but errors, retries and other features can consume more credit. Download the result and keep the transcript/model record before changing settings.

Can the recordings be used outside Inworld?

On-Demand lists a commercial license, and the general and voice-specific terms retain or assign output rights to the user and permit downloaded output outside the service subject to their conditions. Our samples use original scripts and Inworld-designed preset voices; no voice clone or personal recording upload was needed.

Output permission does not grant ownership of another person's voice, license provider demo recordings or remove restrictions in the terms and acceptable-use policy. General termination terms also require deletion of outputs upon termination. Check current terms before committing a recording to a client project.

When to choose Gemini speech in AI Studio instead

Compare our six original AI Studio recordings if you want another browser route with different preset voices and languages.

The Gemini listing also covers a separate developer API. Its tested browser access and preview model limits do not establish a universal Free API allowance or identical output rights.