Skip to main content

Nothing happens when I press the Command Mode shortcut

You press the Command Mode shortcut (Option + Shift on Mac, Alt + Shift on Windows) and nothing happens: no sound, no indicator, no error. Command Mode is off by default, and the shortcut does nothing while it is off. The same applies to the idle pill’s command button and mouse-button bindings.
1

Open Command Mode

Click Command Mode in the sidebar.
2

Turn on Command Mode

Tick the Command Mode toggle at the top of the page.
If there is no toggle, no AI provider is connected. The page shows a locked card with Connect an AI provider. Connect one under Language models. See Set up a language model. If the toggle is greyed out, you’ve reached the Free plan’s daily Command Mode limit. See Daily limits.

AI Enhancement failed, but my text was inserted

An AI Enhancement Failed banner appears, ending in “Processed as-is, no AI enhancement was applied.” The AI Enhancement Failed banner with Switch to provider chips and Hide for 1hr Your dictation is not lost. When enhancement fails, Vowen inserts the unenhanced transcript instead.
  • Switch to: lists your other connected providers. Click one to make it your default and re-enhance the same dictation. While it runs you see “Switching to” followed by the provider name, then “Switched default provider. Enhanced text on clipboard — paste to replace.” Paste to replace the raw text.
  • The chips are hidden when a Tone chose the provider, or when no other provider is connected.
  • Hide for 1hr hides this error for an hour.
Vowen also counts these as failures: an empty reply (usually a reasoning model that never answered), and a reply addressed to you instead of your cleaned text.

Command Mode failed

A Command Mode Failed banner appears after a command. The Command Mode Failed banner with a Switch to row of other providers
  1. Retry with another provider from the Switch to: row.
  2. Hold the shortcut for the whole command, or set it to Tap in Settings › Shortcuts.
  3. Use a model that follows instructions well, such as Claude Sonnet 4.6, GPT-5.4 or Gemini 3.8 Flash.
  4. If the command needed your screen, check that Context-awareness is on, on the Command Mode page. It is off by default.
  5. Connector tools are read-only. Command Mode can look things up in connected apps but can’t change anything in them.

The model can’t read images

You see “This model can’t read images. Turn off Context-awareness, or pick a model that can read images in Language models.”, or the result banner says Screen not shared. Your model only accepts text, so Vowen sends the command without the screenshot. Commands on selected text still work. To include the screen:
  • Pick a model that reads images, such as Gemini or a GPT-5.4 model.
  • On Azure or Custom API, turn on Accepts images if your model supports images.
The Custom API panel with a Default model marked Text only and the Accepts images switch If a model rejected an image once, Vowen treats it as text only for 30 days. Saving the provider again on Language models resets that.

API key not working

You see “Invalid API key. Please check your API key and try again.”, “API quota exceeded”, or a network error when you click Connect.
  1. Copy the key again from the provider’s dashboard. Check for spaces at either end.
  2. Confirm the account is active. Some providers need a payment method even for free use.
  3. “API quota exceeded” means the plan’s usage limit was reached. Check billing with the provider, or try another model.
  4. For a network error, check your connection. A VPN, proxy, firewall or antivirus that scans HTTPS traffic can block the request.
  5. If requests fail instantly on a network that advertises IPv6 without routing it, keep Settings › Developer › Prefer IPv4 Connections on.

Custom model ID doesn’t connect

The ID must match the provider’s docs exactly, including any prefix such as openai/ on Groq. Vowen tests it when you click Connect.

AWS Bedrock doesn’t list a model

Some models, including newer Claude models, are only available through inference profiles. Vowen lists your account’s system-defined inference profiles alongside its models. Your key needs permission to list inference profiles (bedrock:ListInferenceProfiles).

AI Enhancement is slow

Vowen already asks each model for the least reasoning it accepts. There is no setting for it. Groq and Cerebras are built for low latency.

Rate limit errors

Enhancement works for a few dictations, then fails with a rate limit or quota error.
  • Wait. Most provider limits reset within a few minutes.
  • Configure a second provider and use the Switch to: chips.
  • Turn off enhancement for quick dictations, globally or with an alternative shortcut set to Off.
  • Move to a paid tier with the provider.

Daily limits

The Command Mode daily limit is a Vowen limit, not a provider limit. When you reach it you see “Daily Command Mode limit reached”. Free plans get 5 runs a day; free plans from before the current pricing get 20. The Command Mode page shows how many you’ve used today. Pro removes the limit.

AI ignores my instructions

  1. Keep instructions short and concrete.
  2. Check that two instructions don’t contradict each other.
  3. If a Tone is active, check its instructions too. A Tone can change the provider and add its own rules.
  4. Use a more capable model. Small, fast models follow fewer rules.
Global Custom Instructions apply to AI Enhancement only. Command Mode uses the active Tone’s instructions instead. See Custom Instructions.

The AI rewrites my sentences or adds words

Enhancement should only fix errors. Smaller open-weight models sometimes paraphrase or add things anyway.
  1. Switch to a larger model, such as Claude Sonnet 4.6, GPT-5.4 or Gemini 3.8 Flash.
  2. Add a custom instruction such as “Do not rephrase. Only fix errors.”
  3. If the transcript itself was wrong, fix the speech side first. See Transcription quality.

Model names don’t match what I read elsewhere

Open Language models, click the provider’s gear, and read the Default model dropdown. That list is current: Vowen updates it from an online catalog. One model per provider is marked (Recommended). To use a model that isn’t listed, choose Custom model ID….