How to Install Text-to-Speech Models in Chrome

How to Install Text-to-Speech Models in Chrome

You install a text-to-speech model in Chrome by downloading the voice engine onto this computer, then choosing it in Voice settings. In Flow TTS that work lives in Settings → Models. A model is the engine. A voice is how that engine sounds. You do not send each page to a cloud TTS API to try a new family — you install the files first, then play. This is not the same question as what makes a voice sound natural. That post is quality. This one is the Models tab.

Do you need to install a model to start?

No. Flow TTS ships with Supertonic v1 already bundled. You can get started and press play without opening Models.

Every other catalog entry must be installed before you play it. The Models tab says Install required on those, and the detail panel repeats that the model must be installed before playback. Supertonic v1 can show as Bundled (not managed) until you treat it like a downloaded model.

The registry currently lists 27 model IDs across five families: Supertonic, Kitten, Kokoro, Piper, and TinyTTS. The left column groups them by family. Pick one row to see its detail on the right.

How do you install a model?

1. Open Flow TTS Settings.

2. Open the Models tab.

3. Choose a family, then a specific model.

4. Click Install.

5. Wait until the status is Complete, not only Installed.

The panel shows a progress bar, bytes downloaded, and a short phase line while the job runs. You can Cancel a download that is queued or installing. If the connection drops, use Retry. When the error mentions resume, Retry continues from the saved checkpoint instead of starting over.

Update replaces an installed model with a newer package when one is available. Remove deletes the downloaded files from this browser profile. Refresh reloads the list if something looks stale.

Downloads use the network. Files come from each model’s listed source — often a public model host — and Flow TTS checks them as they arrive. Installed files stay on this computer for the next session. That is still not fully offline: account, usage, and later downloads can need the network again.

What do the status labels mean?

Read the badge before you switch voices.

• Install required: this model is not on disk yet. Play will fail if you select it now.
• Bundled (not managed): Supertonic v1 is available without a Models-tab install.
• Queued / Installing: the download is in progress. Leave the tab open or come back; you can Cancel.
• Installed plus Complete: required files are present. Safe to choose in Settings → Voice.
• Incomplete: the row is marked ready but a required file is missing. Use Update or Remove and Install again.
• Needs attention: something failed. Read the error, then Retry or Remove.

Each model also shows a Performance label: light, balanced, or demanding. That is a hardware hint, not a quality score. A demanding model can sound better and still stall on a small machine. A light model starts sooner and may sound thinner. Change one family at a time so you can hear the difference.

What are voice packs?

Some models — Kokoro is the usual case — ship extra voice packs. The engine can be installed while additional voices are still optional. The Models tab lists those packs under the selected model. Install only the voices you will use. They take space and time.

After a pack is on disk, pick the voice in Settings → Voice. Installing a pack does not switch the active voice by itself. Pronunciation rules are separate; they rewrite words after you already have a model that can speak.

Where Flow TTS fits

The Models tab is for people who want a second local engine without leaving Chrome.

Keep these limits in mind:

• Play still needs an email one-time code. The tour can continue unsigned-in; playback cannot.
• Free usage is 10,000 characters per day. Premium unlocks unlimited characters and Glass Mode when entitlement is active.
• Speech synthesis is local for supported installed or bundled models. Accounts, licenses, usage, and model downloads use the network. See the privacy policy.
• Do not remove the only ready model and then expect play to work.

If you only need to hear one article today, stay on bundled Supertonic. Open Models when you have time for a download and a short listen test.

FAQ

How do I install a text-to-speech model in Flow TTS?

Open Settings → Models, select a model that says Install required, click Install, and wait for Complete. Then choose it in Settings → Voice.

Why can’t I play after I clicked Install?

The download may still be running, or the badge may say Incomplete or Needs attention. Stay on bundled Supertonic until the new model is Complete.

Can I pause a model download?

Yes. Cancel stops a queued or installing job. If an error says resume is available, Retry continues from the checkpoint.

Is installing a model the same as going offline?

No. The files live on this computer after a successful install, but account checks, usage, and future downloads can still need the network.