VoicePaste
中文
Documentation
Model API setup

DeepSeek

DeepSeek handles text translation and polishing in VoicePaste, not speech recognition. It is optional for plain transcription.

1. Get an API key

  1. Sign in to the DeepSeek platform.
  2. Open API keys, create a key, and store it.
  3. Check the API account's balance and available services.

2. Configure VoicePaste

  1. Open Settings → Large Language Model and select DeepSeek.
  2. Enter the API key. You can use the app's preset endpoint, https://api.deepseek.com/v1/chat/completions.
  3. Use a model identifier currently supported by the API. Start with the app preset.
  4. Finish editing the fields, then configure a text template.

This field expects the full Chat Completions endpoint, not the website homepage or console URL.

3. Verify translation or polishing

  1. Add a text template, such as “Translate to English,” and write its instruction.
  2. Bind a shortcut to it in Settings → Shortcuts.
  3. Say a sentence in an input field, finish with the template shortcut, and check the translation.

See Language models and templates for the full flow. Saving a key alone does not verify an API call.

Common questions

What if authentication fails?

Check the complete key and which API account owns it.

What if the balance or quota is insufficient?

Check API billing on the platform. Web chat access does not establish that the API has usable quota.

What if the model is not found?

Check and update the API model identifier using DeepSeek's documentation. Do not use a web chat mode's display name.

What if recognition works but polishing fails?

Recognition and DeepSeek are separate stages; check their settings separately.

On this page