Text-to-Speech in ImageLayer: Current Models, Voices, and Quotes
A current guide to ImageLayer's public Google Gemini text-to-speech models, compatible voices, request quotes, and safe review workflow.
Product status — October 14, 2026: This guide describes the current public model catalog. The scripts below are fictional teaching examples, not customer work or evidence of a particular result. Model availability and the request quote shown in the widget are the source of truth when you submit.
ImageLayer’s Audio mode turns a written script into standalone speech. Choose a public TTS model, select a voice that is compatible with that provider, choose MP3 or WAV, and review the request-specific quote before generating.
Current public TTS models
The current model manifest offers two public audio models:
| Model shown in ImageLayer | Provider | Compatible voices | ImageLayer launch capability |
|---|---|---|---|
| Gemini TTS Flash | Kore, Puck, Charon, Aoede, Sulafat, Achird, Leda, Sadaltager | Single-speaker only; the launch manifest intentionally does not expose multi-speaker output, even where provider documentation describes it | |
| Gemini TTS Pro | Kore, Puck, Charon, Aoede, Sulafat, Achird, Leda, Sadaltager | Single-speaker only; the launch manifest intentionally does not expose multi-speaker output, even where provider documentation describes it |
Model names such as “Flash” and “Pro” are provider labels. They are not a promise that one model will sound better, finish faster, or fit every script. Compare the models with your own approved script and listening criteria.
Current voice options
The two current models use eight Google voices. These style words are selector labels, not measured quality ratings.
| Voice | Selector style |
|---|---|
| Kore | Firm |
| Puck | Upbeat |
| Charon | Informative |
| Aoede | Breezy |
| Sulafat | Warm |
| Achird | Friendly |
| Leda | Youthful |
| Sadaltager | Knowledgeable |
Select the model first, then choose from the compatible voices shown in the widget. Google speech requests can apply the supported tone instruction shown by the interface.
OpenAI TTS and TTS HD are no longer offered for new requests after their provider announced deprecation. Previously generated files remain available under their existing retention policy. A replacement voice model will appear only after its integration is verified.
How the quote works
Audio does not have one universal price per sample. ImageLayer’s pricing calculator evaluates the selected model, requested duration, and the exact script size. Longer requests can add credits, and the calculator can raise the quote when required to preserve the configured provider-cost margin.
Treat the quote displayed for the finished request as authoritative. Do not reuse a credit number from a screenshot, article, or earlier run.
Walkthrough: a fictional product narration
-
Open Audio mode.
-
Select Gemini TTS Flash and Kore, or choose another currently available model-and-voice pair.
-
Keep MP3 for a compact standalone export, or choose WAV when your downstream workflow requires it.
-
Paste a bounded script such as this:
Prism is a fictional smart-lamp concept created for this ImageLayer tutorial. This sample narration introduces its brushed-aluminum form and adjustable ambient light without making claims about a real product or customer.
-
Review the quote shown for that exact request before submitting.
-
Listen to the entire result. Check pronunciation, pauses, missing words, clipped endings, and whether the delivery matches the written brief before publishing it.
No audio file is embedded in this walkthrough. A public sample should be added only after its bytes, complete transcript, duration, model, voice, and review status are recorded together.
Write scripts that are easier to review
- Use ordinary spoken sentences. Read the script aloud before synthesis and shorten clauses that are hard to say in one breath.
- Define abbreviations. Spell out product codes or acronyms when pronunciation matters.
- Write numbers for the intended reading. For example, choose between “twenty-five percent” and “two five percent” deliberately.
- Keep factual claims sourced. TTS makes the words audible; it does not validate them.
- Review the whole export. A playable file can still contain a mispronunciation, unsupported claim, or clipped final phrase.
Availability
Audio generation depends on the plan, organization settings, admitted model catalog, and provider readiness for the signed-in organization. The model selector and pre-submit quote reflect the current request; this article does not override either one.
Keep exploring
- Prepare Podcast-Style TTS Snippets Safely — A launch-safe monologue workflow and the current status of multi-speaker output.
- Audio & Text-to-Speech Guide — A shorter in-product walkthrough.
Inspect the current workflow
Open the playground to inspect the Audio interface and configuration flow. The public preview uses sample data and does not call a generation provider. For an actual standalone single-speaker request, sign in to an eligible workspace and rely on its live admission, quote, credits, spend controls, and provider checks. Review the complete file before treating it as publishable.