Read Aloud (Text-to-Speech) Settings in Falcon PDF
Set up Read Aloud with Device TTS, ElevenLabs, OpenAI/OpenRouter, or a self-hosted LocalAI endpoint. Includes per-platform install steps for macOS, Windows, and Linux.
Read Aloudturns any PDF or eBook into spoken audio. Falcon PDF supports several text-to-speech (TTS) providers so you can pick the trade-off that fits you: zero setup with your device's built-in voice, premium cloud voices, or a fully private self-hosted model running on your own machine. This guide walks through each provider and how to set up a local endpoint.
Device TTS and self-hosted LocalAI never send your document text off your machine. ElevenLabs and OpenAI/OpenRouter send the text you read aloud to that provider using your API key, which Falcon PDF stores only on your device.
Step 1: Open Read Aloud settings
- Open any PDF or eBook in Falcon PDF.
- Open the Read Aloud sheet (the headphones / speaker icon), then tap the gear ⚙️. You can also reach it from Settings → Read Aloud (TTS) — a full screen on mobile, a modal dialog on desktop.
- At the top you'll see a row of provider tab-cards: System, ElevenLabs, OpenAI / Gateway, and Local AI. Tapping a card selects that provider and swaps the panel below it.
Step 2: Choose a provider
Pick the provider below to see its settings. Each one has its own panel in the app, mirrored here.
Device TTS (System)
The free, built-in option. Uses the operating system's own speech engine, so there's nothing to install and no API key.
- Select the panel: choose System.
- Pick a language: choose the language/locale first. This filters the voice list to just that locale.
- Pick a voice: the available voices for the selected locale (installed on your device) are shown — pick one.
- Test: tap Test to hear a short preview.
Want more or better system voices? Install extra language/voice packs in your OS settings (Android: Settings → Accessibility → Text-to-speech; Windows: Settings → Time & language → Speech; macOS/iOS: Accessibility → Spoken Content → Voices), then reopen Falcon PDF.
- No voices listed: install a voice/language pack in your OS settings, then reopen Falcon PDF.
- Robotic or wrong-language voice:pick a voice that matches the document's language from the voice list.
- Playback stutters on long text: lower the playback speed.
ElevenLabs
Premium, highly natural cloud voices. Requires an ElevenLabs account and API key.
- Create a key at elevenlabs.io → Settings → API Keys.
- Choose the ElevenLabs panel, paste the key, then tap Validate.
- Pick a voice from the voice picker and tap Test.
ElevenLabs bills per character of synthesized audio. Long documents can use your quota quickly — check your plan's monthly character limit.
- Validation fails: re-copy the key (no stray spaces or line breaks) and confirm your ElevenLabs account is active.
- Quota / 401 errors:you may be out of characters for the month — check your plan's character limit.
- Playback stutters on long text: lower the playback speed.
OpenAI / OpenRouter (Gateway)
Natural AI voices through OpenAI's TTS API, OpenRouter, or any OpenAI-compatible gateway. A preset dropdown pre-fills the endpoint; you supply the key, model, voice, and output format.
| Preset | Endpoint (full speech URL) | Format |
|---|---|---|
| OpenAI | https://api.openai.com/v1/audio/speech | wav |
| OpenRouter | https://openrouter.ai/api/v1/audio/speech | pcm |
| Custom | your gateway's /v1/audio/speech | wav |
- Get a key: platform.openai.com/api-keys or openrouter.ai/keys.
- Select the preset — the endpoint fills in automatically.
- Paste the API key and tap Validate.
- Set the model, voice, and output format (see the options below).
- Tap Test, then Save.
| Field | Common values | Where to find them |
|---|---|---|
| Model | tts-1, tts-1-hd, gpt-4o-mini-tts | OpenAI TTS models |
| Voice | alloy, echo, fable, onyx, nova, shimmer | OpenAI voice options |
| Format | wav, mp3, pcm | Speech API reference |
There is no "list voices" endpoint in the OpenAI TTS API, so you type the voice id. The values above are OpenAI's. For OpenRouter or another gateway, use that provider's documented model and voice names — browse the OpenRouter speech models.
OpenRouter lists some speech models with a free tier, so you can get text-to-speech at no cost. Filter for them on the speech models page (look for :free variants), use the OpenRouter preset, and paste your free OpenRouter key.
Not sure which model, voice, and format to use? Paste a prompt like this into ChatGPT, Gemini, or any AI chatbot to get a quick comparison table to try:
List the text-to-speech (TTS) models available on OpenRouter.
Return a Markdown table with these columns:
Model ID | Available voices | Supported output formats | Free or paid
Only include models that output speech/audio.Swap "OpenRouter" for "OpenAI" to list OpenAI's TTS options instead. Then enter a row's values into the model, voice, and format fields and tap Test.
- Validation error: re-check the key, the exact model name, and that the endpoint ends in
/audio/speech. - No audio / empty response: the voice id may not exist for that model — use a documented voice name.
- Distorted audio on OpenRouter: keep the output format as
pcm(the app assumes 24kHz mono for that preset).
LocalAI (self-hosted, private)
Run a neural TTS model on your own computer and point Falcon PDF at it. Nothing leaves your machine, there are no usage bills, and it works offline once set up. The app talks to any OpenAI-compatible speech server, so you can use LocalAI, Kokoro-FastAPI, or any similar server.
- Start a local speech server on your machine (see Setup local Endpoint below).
- Choose the Local AIpanel and set the endpoint to your server's full speech URL (e.g.
http://localhost:8080/v1/audio/speech). - Tap Check Daemon to confirm the server is reachable.
- Tap Add Local Model / Voice and enter a voice identifier (e.g.
af_bella,am_adam), a locale, and gender. - Use Set Active + Test on a registered voice to hear it, then Save.
Setup local Endpoint
Pick a local server to run. Each project has its own cross-platform install guide (macOS, Windows, Linux) — use the server's own docs for the exact steps on your OS. We give full install steps for LocalAI below.
LocalAI is an OpenAI-compatible server that exposes /v1/audio/speech (and /v1/chat/completions) on port 8080. For OS-specific installers and all run options, see the LocalAI installation guide.
Quickest start (Docker):
docker run -p 8080:8080 localai/localai:latest- In the Read Aloud Local AI panel, set the endpoint to the full speech URL:
http://localhost:8080/v1/audio/speech- Leave the API key blank (local endpoints usually don't require auth), then tap Check Daemon.
- Tap Add Local Model / Voice, register a voice id your LocalAI model supports, then Set Active + Test.
- Reuse the same server for AI Chat by pointing the Custom Provider base URL at
http://localhost:8080/v1.
Any OpenAI-compatible speech server works. Follow its repo's own install instructions for your platform, then paste its full /v1/audio/speech URL into the Local AI panel.
- Kokoro-FastAPI — Kokoro voices, default port
8880. See the repo's README for macOS / Windows / Linux setup. - LocalAI — see the LocalAI tab for full steps.
Bind the server to 127.0.0.1 (localhost) only, not 0.0.0.0 — a TTS server with no auth must not be reachable from outside your machine. The endpoint field is used exactly as entered (trailing slashes trimmed), so a non-standard speech path still works. If a port is taken, run the server on another port and update the endpoint to match.
- "Check Daemon" fails: confirm the server is running and the port matches. The
/v1/audio/speechpath isn't a GET route, so check the server's own health/docs URL in a browser (e.g.http://localhost:8880/docs). - No audio / empty response: the voice id may not exist on the server — use a voice the model ships (e.g.
af_bellafor Kokoro). - Playback stutters on long text: lower the playback speed or switch output format (WAV is the safest, self-describing format).
- Firewall prompt on first run: allow the local server on the loopback interface only.
FAQ
Which provider should I pick? Device TTS for zero setup, ElevenLabs for the most natural cloud voice, OpenAI/OpenRouter if you already have a key, and LocalAI for full privacy and offline use.
Does LocalAI cost anything? No — it runs on your own hardware. You only pay the one-time effort of installing the server.
Is the local option available on mobile? Self-hosted LocalAI is a desktop feature (macOS, Windows, Linux). On mobile, use Device TTS, ElevenLabs, or OpenAI/OpenRouter.
What next? Pair Read Aloud with your own AI providerto chat with the same document you're listening to.
Related articles
How to Configure Your Own AI Provider in Falcon PDF
Connect OpenAI, OpenRouter, Groq, Together AI, or a local Ollama model and chat privately with your PDFs using your own API key.
6 min readPDF EditingHow to Edit a PDF: A Complete Guide
Edit any PDF like a pro — change text, add images, rearrange pages, fill forms and annotate — on iOS, Android, Mac, Windows and Linux with Falcon PDF.
8 min readPDF EditingHow to Add Text to a PDF
Type directly onto any PDF — fill flat forms, add notes, insert labels and match fonts — on iOS, Android, Mac, Windows and Linux with Falcon PDF.
5 min read