You can listen to AI replies in different languages on your iPhone without sending the text to a cloud speech service. OGAM (Off Grid AI Mobile) Pro generates speech on the phone. Download the local voice model and the voices you need first. Then you can play supported text with Wi-Fi and mobile data off.
Get OGAM on the App Store | Mobile features
What would you like to do with Off Grid AI?
Have a feature or use case you would like us to support? Tell us what you want to do and which device you use.
Write to support@offgridmobileai.co, join our Slack community, or talk to us on Reddit.
This guide covers speech playback for assistant messages inside OGAM. It does not set up a voice for every app on your iPhone.
For example, you can ask for a short reply in French, check the text, then hear it with a French voice. Changing the voice language does not translate the reply. The text needs to be in the language you want to hear.
What do you need for offline speech on iPhone?
You need a supported iPhone, OGAM Pro, and the local voice assets downloaded on the phone. The app's published requirements include iPhone 12 or newer on iOS 17 or later. Finish app installation, Pro setup, and the voice downloads while connected.
Prepare these items:
- The current App Store version of OGAM with Pro access.
- Free storage for the speech model and selected language assets.
- A completed assistant reply to play aloud.
- A local chat model if you also want to generate new replies offline.
Text-to-speech is a Pro feature on mobile. Speech recognition, which turns your voice into text, is a different feature. This article uses typed text and speech output, so you do not need to record your voice.
The mobile feature page lists the platform requirements. GitHub and App Store releases can differ. Use the settings available in your installed iPhone version and update the app if the language controls are missing.
Which languages can you use?
Choose from the Language menu in the text-to-speech settings. The app builds that menu from the voices supported by its active speech engine. French, Spanish, Hindi, and English voices are examples in the local voice system; use the actual menu in your installed app as the available list.
Select a language first, then a Voice within that language. If changing the language starts a download, wait for it to finish before going offline.
A voice that is ready in English does not mean every other language is already downloaded. The app caches the assets it needs for the selected voice. Changing languages can require additional files.
Speech recognition and speech output have different language menus. The languages supported by a transcription model do not establish which voices are available for reading text aloud.
How do you download a local voice model?
Open Models > Voice and download the local voice model. The model is listed as Kokoro TTS in the current implementation. Wait for the download and preparation to finish. Select the local voice model if a remote voice service was previously active.
The model name helps you find the right entry; you do not need to install a separate speech engine or use an API key.
Keep the phone connected for this setup. Do not treat a model card appearing in the list as proof that its files are already on the device.
If you see a message that no voice model is downloaded, return to Models > Voice. Some settings can refer to TTS Settings, but the model download lives in the Voice tab of Models.
How do you choose the language and voice?
Open Model Settings > Text to Speech. Use Chat mode and turn on Enable TTS. Then select Language and choose a Voice. Wait for the message that the selected language's voice is ready. These settings control speech playback for assistant messages.
You can also find the text-to-speech section in Chat Settings.
For your first check:
- Select a language you can understand.
- Choose one voice from that language's list.
- Wait for any download or preparation to complete.
- Leave Speed at 1.0x until you have heard the result.
A failed language download has a Retry action. Retry while connected. Switching airplane mode on before the new voice is ready will not finish that download.
How do you play a reply aloud without internet?
After setup, use the speaker button on a completed assistant message. OGAM generates the audio using the selected local voice. Press the active speaker button again to stop playback. For a full offline check, turn on airplane mode and make sure Wi-Fi is off before playing the message.
1. Prepare a short reply
With a local chat model selected, ask for a short sentence in your chosen language. For French, you could ask:
Reply in French with one sentence saying that the meeting starts on Monday at ten.
Wait until the assistant's text is complete. Read it first. The speaker button does not establish that the model's sentence is accurate or that it followed your request.
2. Match the speech language
Select French in the text-to-speech Language setting and choose a French voice. Wait for the voice-ready message.
Use a different supported language if you prefer. Match the text and voice language for the first test.
3. Disconnect and play
Turn on airplane mode. Check Wi-Fi is off too. Return to the completed reply and tap its speaker button.
You should hear the reply without needing to send another prompt. This tests speech playback separately from chat generation.
4. Check a second language
Reconnect to prepare another language if its assets are missing. Choose a voice, wait until it is ready, and prepare a matching reply. Repeat the disconnected playback check.
Keep each test short. You can assess pronunciation and volume without waiting for a long passage.
Can it read your own text exactly?
The verified playback control reads assistant messages. If you ask a local chat model to repeat your text, check its reply before playing it. A chat model can change punctuation, omit words, or rewrite a sentence even when you request an exact copy.
For example, ask it to repeat a short passage without additions, then compare the reply with your original. Do not treat this as a guaranteed verbatim document reader.
This workflow also does not automatically translate a paragraph. Ask the chat model for a translation first, review it, then select a voice for the target language.
What should you check if the voice does not play?
| Problem | What to check | Action |
|---|---|---|
| No speaker button appears | Pro access, Enable TTS, model readiness, reply completion | Complete setup, enable TTS in Chat mode, and wait for the assistant reply to finish. |
| A new language fails offline | Its voice assets may not be downloaded | Reconnect, select that language, wait for the ready message, then test offline again. |
| The sound comes from another device | iPhone volume and audio output | Check media volume and the connected headphones or Bluetooth speaker. |
| Pronunciation is wrong | Text language and selected voice | Match them, then try a short sentence with ordinary words. |
| Speech works but a new reply fails offline | The chat model may be remote | Select a downloaded local chat model. |
| The settings show a remote voice service | Active voice model | Return to Models > Voice and select the local model. |
| Voice preparation fails | Download completion and available memory | Finish the download, close memory-heavy apps, and try a short reply. |
Names, abbreviations, and mixed-language sentences can sound different from what you expect. A voice that pronounces one sentence well is not a guarantee for every passage.
Does the text leave your iPhone?
With the local voice model selected, speech synthesis runs on your iPhone. A remote voice service uses a different path and can send the text to that service. Keep the selected voice local for the workflow described here.
If you generate new replies, also check the chat model. Local speech playback can read text that was produced by a remote model. That does not make the earlier chat request local.
iPhone text-to-speech FAQ
Is multilingual speech output free?
Text-to-speech is part of OGAM Pro. Downloading the app for free does not include every Pro feature. Check the app's current Pro terms before purchase.
Does changing Language translate the message?
No. It selects the language and voice used to speak the text. Translation requires a separate step.
Do I need mobile data after downloading a voice?
The local speech-processing step does not need mobile data once the required assets are available. Finish Pro setup first and check playback with airplane mode enabled.
Can I use this voice in every iPhone app?
This guide sets up OGAM's assistant-message playback. It does not replace the iPhone's system voices or add speech controls to other apps.
Prepare one voice, play one completed reply offline, then download the other languages you want before you travel.
Top comments (0)