ElevenLabs voice tool: batch MP3 on your computer
Set up the connection once, test a difficult Vietnamese paragraph, then process a folder of source files into MP3.
tofu-voice-EL is the Tofu Voice edition for an ElevenLabs desktop workflow. The current app bundles 756 voices across 21 language codes and can process long files or folders instead of requiring you to paste every segment into a web box.
Who should use tofu-voice-EL
- Audiobook and story creators with many chapters and a structured output folder.
- Creators who want to compare Vietnamese voices and repeat the same desktop workflow.
- People working with TXT, DOCX, EPUB, text-based PDF, Markdown, HTML, SRT, or VTT.
- Users prepared to configure a compatible proxy for the ElevenLabs lane.
If you first need to evaluate pronunciation, read ElevenLabs Vietnamese voice testing before choosing an edition.
Prepare the first generation
- Install tofu-voice-EL and enter its `TFVE` activation key.
- Open the ElevenLabs voice family. Add a TopProxy key or enter a compatible proxy in the fields shown by the app.
- Check each connection and continue only when its status is working.
- Create separate source and MP3 output folders so the original documents are never overwritten.
Choose a Vietnamese voice and create MP3
- Select Vietnam/Vietnamese to narrow the catalog.
- Preview a few voices without starting a long job yet.
- Test: “Hải hỏi: Bạn đã sửa kỹ chương bảy chưa? Nếu ổn, mình bắt đầu lúc tám giờ rưỡi nhé.”
- Choose a file or folder, confirm the MP3 destination, and export one representative file.
- Listen to the beginning, middle, and end. Run the batch only after names, numbers, and pacing pass.
Supported source formats
- TXT works best for a story already split by chapter.
- DOCX, EPUB, Markdown, and HTML preserve a document-based workflow.
- PDF must contain selectable text; a scan or image needs OCR first.
- For SRT and VTT, the ElevenLabs lane creates continuous speech and does not preserve video timestamps.
Batch files without repeating finished work
Name files `01`, `02`, and `03`, and keep each MP3 name aligned with its source. If the connection stops, inspect the output folder and restart from the first missing file. The multiple TXT audio guide has the full batch and resume checklist.
- Keep the computer awake during online generation.
- Do not switch voices mid-batch when you need a consistent narrator.
- Spot-check a chapter in the middle, not only the first output.
- Keep the source files and export MP3 into a separate folder.
Base, Pro, and ElevenLabs editions
- Base: TXT with Edge and CapCut voices.
- Pro: adds documents, subtitles, offline voices, and local voice clone; no ElevenLabs lane.
- ElevenLabs: includes the Pro tools plus the ElevenLabs lane; that lane needs a proxy.
Three common failures
The proxy is not ready
Check the SOCKS5 and HTTP ports, proxy credentials, and the status of each key. Do not start a batch while the app still asks you to verify the connection.
A pleasant voice mispronounces Vietnamese
Choose a voice with a Vietnamese sample. Spell out numbers, symbols, and abbreviations as you want them spoken. Do not process the full work while the test line is still wrong.
A PDF produces empty or missing speech
Open the PDF and try to select text. If you cannot, it is probably a scan and needs OCR before text-to-speech.
FAQ
Are all 756 voices Vietnamese?
No. 756 is the current total bundled catalog across 21 language codes. Filter Vietnamese and test your real text.
Does buying the app make ElevenLabs unlimited?
The app does not impose a total character quota on the file workflow, but the ElevenLabs lane still depends on the proxy and upstream service. A one-time app purchase is not a promise that upstream access is permanently free or unlimited.
Can one key run on two computers?
No. One key has one active computer. Activating another computer replaces the current one.
How much does tofu-voice-EL cost?
Check the current price and availability on the Tofu Voice product page because the catalog can change over time.