Feature Request: Support for self-hosted / local TTS in Read Aloud
Zotero Read Aloud would benefit substantially from native support for self-hosted text-to-speech (TTS) engines, particularly lightweight neural models like Kokoro-82M. Adding a configuration field for an OpenAI-compatible speech endpoint (`/v1/audio/speech`) would allow users to connect local TTS servers directly to the built-in reader.
Local neural TTS addresses critical gaps in the current voice offerings. System voices run offline but lack natural prosody and cadence, making sustained academic listening difficult. Zotero cloud voices sound excellent but require constant network connectivity, consume monthly quota allowances, and transmit document text to external servers. Many researchers process confidential manuscripts, peer reviews, clinical data, or proprietary archives that institutional compliance rules prohibit transmitting off-device.
The configuration requires few controls in Zotero's Read Aloud preferences. Adding a "Custom / Self-Hosted" option under voice selection would require three settings:
1. **Endpoint URL** (e.g., `http://localhost:8880/v1/audio/speech`)
2. **Voice Identifier** (e.g., `af_heart`, `am_adam`)
3. **API Key / Bearer Token** (optional, for authenticated local or LAN reverse proxies)
This addition would give privacy-conscious researchers access to high-fidelity narration while preserving Zotero's native sentence-level tracking.
Local neural TTS addresses critical gaps in the current voice offerings. System voices run offline but lack natural prosody and cadence, making sustained academic listening difficult. Zotero cloud voices sound excellent but require constant network connectivity, consume monthly quota allowances, and transmit document text to external servers. Many researchers process confidential manuscripts, peer reviews, clinical data, or proprietary archives that institutional compliance rules prohibit transmitting off-device.
The configuration requires few controls in Zotero's Read Aloud preferences. Adding a "Custom / Self-Hosted" option under voice selection would require three settings:
1. **Endpoint URL** (e.g., `http://localhost:8880/v1/audio/speech`)
2. **Voice Identifier** (e.g., `af_heart`, `am_adam`)
3. **API Key / Bearer Token** (optional, for authenticated local or LAN reverse proxies)
This addition would give privacy-conscious researchers access to high-fidelity narration while preserving Zotero's native sentence-level tracking.
-
AbeJellinek Zotero TeamThis is planned.
-
AbeJellinek Zotero Teamedited yesterday at 4:28pm(But note that Zotero's Standard voices do not transmit text outside Zotero infrastructure, and you get unlimited minutes with a storage subscription. See the available voice types and quotas for your account in your account settings.)
Upgrade Storage