Voice mode: audio playback does not auto-start on the reply following a pasted document
Description
In a voice-mode conversation, when the user pastes a document or a large block of text into the chat input instead of speaking, Claude's next reply is rendered as text but the spoken audio does not begin playing automatically the way it does for replies to spoken turns. The user has to intervene to hear the response, which breaks the hands-free flow that voice mode exists for. This is especially disruptive for accessibility-driven use: the reporter works by voice from a phone and relies on replies being read aloud without touching the screen, so every pasted artifact silently stalls the session until noticed.
Steps to Reproduce
1. Start a voice-mode conversation with Claude. 2. Confirm normal behavior: speak a turn, and observe that Claude's reply is spoken aloud automatically. 3. Instead of speaking, paste a document or a long block of text into the chat input and send it. 4. Observe Claude's reply to the pasted content.
Expected Behavior
Claude's reply is spoken aloud automatically, the same as replies to spoken turns, regardless of whether the preceding user turn was spoken or pasted.
Actual Behavior
The reply appears as text but audio playback does not auto-start. No error message is shown; the session simply goes silent until the user manually triggers playback.
Environment
Product
Claude voice mode
Environment
Claude voice mode, mobile phone client. Long-running voice session mixing spoken turns with pasted text artifacts.
Activity
No comments yet.
