Choose how to create
- Remote MCP
- Connect Narrivela to your AI client with the HTTPS server URL and your Narrivela account.
- Web studio
- Use the studio to sign in and generate speech in your browser.
- Remote MCP
- Connect with MCP in a supported AI client using your Narrivela account.
- Client guides
- Codex · ChatGPT · more clients
First-release voices and languages
The first public release uses Qwen Audio 3.0 TTS Plus (Beijing) for ordinary narration in Mandarin Chinese and English, with two bilingual voices: female longanlingxin and male longanlufeng. Style and emotion prompts are not available in this release.
The examples page provides short Mandarin Chinese and English samples of the current female voice. Listen to your own script before a large batch. Actual input and queue limits are shown by the studio.
The web studio workflow
- Write or upload your scripts
Use single-script mode for a complete narration or batch mode for scripts separated by blank lines. UTF-8 text and JSONL input are supported.
- Choose the sound
Select an available female or male voice, Mandarin Chinese or English, speed, and output format. Style prompts are not available in the first public release.
- Generate, listen, and save
Check task progress, listen to clips, and download WAV, MP3, or a ZIP. After a refresh, the studio attempts to recover the most recent task.
Settings and output
- Voices and languages
- Narrivela offers two bilingual voices for Mandarin Chinese and English. Select Mandarin Chinese or English when submitting a task.
- Speed
- The studio shows the available speed range. Listen to check names, numbers, and pauses.
- Style and emotion
- Style and emotion prompts are not available. Choose the available voice and speed settings instead.
- WAV and MP3
- WAV is useful for further audio processing. MP3 is convenient for listening and distribution.
- Batch ZIP
- Clips and a manifest follow the original submission order. Large results can be viewed in pages.
- Errors and recovery
- Fix invalid input before resubmitting. Try later when the queue is full. Resolve provider or network errors before resuming the original task.
Scripts and results
The server sends the script to its configured speech provider. Production retention periods and deletion rules are described in the privacy notes. Completed audio is normally available for about 24 hours; download it before the displayed expiry time.
Connect through ChatGPT
Check ChatGPT installation, authorization, and prompt examples.
Connect through Codex
Other MCP clients
Connect Narrivela with MCP in WorkBuddy, Claude, Cursor, VS Code / Copilot, Gemini CLI, Dify, n8n, or another HTTP MCP client. The guide includes the server URL, sign-in steps, example configurations, speech tools, and downloads.