Current walkthrough — August 25, 2026: The script below describes a fictional dashboard concept. It is not customer work or evidence of product performance. The signed-in widget is the source of truth for model availability and the request quote.
Audio mode turns a written script into standalone speech. Keep the first walkthrough short, concrete, and easy to review from the first word through the final silence.
What you’ll learn
- How to choose a current model and a compatible voice
- How MP3 and WAV change the exported file type
- How to review the quote for the exact request
- How to validate the complete result before publishing it
Before you start
- Open Generate
- Switch to Audio mode
- Use only models and voices shown as available for your organization
- Keep the first demo on
MP3with one short narration script
Choose Gemini TTS Flash, Kore, and MP3
Click Audio in the mode selector, then choose the settings below if they are enabled for your organization.
- Under
Model, chooseGemini TTS Flash. - Under
Voice, chooseKore - Firm. - Under
Format, keep the first demo onMP3. - The voice style is a selector label, not a guarantee about the generated delivery.
- Switch to
WAVonly when your downstream workflow requires it.
Expected result
The request uses a current public model, a compatible Google voice, and a defined export format.
Paste one short, explicitly fictional narration script
Paste this script into the textarea:
This is a fictional dashboard walkthrough created for an ImageLayer tutorial. Review the visible campaign summary, compare the displayed metrics, and choose which draft to refine next.- Write in full sentences, not fragmented notes.
- Keep the first script to one or two short lines.
- Use punctuation deliberately, then verify the generated pauses by listening.
- Keep
Gemini TTS Flashselected for this walkthrough. - Review the credit quote for this exact script and requested duration before clicking
Generate.
Expected result
The request is bounded, reviewable, and does not turn a fictional example into a customer or performance claim.
Listen through the ending, then keep or revise the read
Once the player appears, listen all the way through before choosing the next action.
- Use
Downloadonly after the complete file passes review. - Use
Generate againwhen the script is correct but the output does not meet the brief. - Use
Edit promptif the wording itself needs work. - If tone is the problem, change
Voicebefore regenerating.
Expected result
The reviewer can catch unsupported wording, pronunciation errors, missing words, or a clipped ending before publication.
No public MP3 is embedded in this guide. Add a sample only when its complete transcript, duration, model, voice, and human playback review are bound to the exact file.
Recommended defaults for team onboarding
- One compatible model-and-voice pair, such as
Gemini TTS FlashwithKore - Firm MP3as the first export format- One short, explicitly fictional script for tutorials and onboarding
- The current pre-submit quote instead of a credit number copied from documentation
What to read next
- Current TTS models and voices for the full public catalog
- Text Generation if you also want to show script creation inside the same widget
- Docs for audio-related widget events and configuration