Repository navigation
berd-call: wait for playback room instead of failing long Siri replies - #405
Closed
johnmatthewtennant wants to merge 1 commit into
Closed
johnmatthewtennant wants to merge 1 commit into
johnmatthewtennant wants to merge 1 commit into
Conversation
Siri renders faster than playback drains, and the native callback failed the whole reply once 60 seconds of audio were pending. It now waits for room when the bound is full and keeps it as a memory cap. Signed-off-by: John Tennant <jtennant@block.xyz> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
Author
|
🤖 Folded into #416 for review and merge. |
johnmatthewtennant
added a commit
that referenced
this pull request
Oct 22, 2026
…rols (#416) ## Summary Several improvements to Berd Call’s speech playback, Siri voices, microphone controls and experimental Realtime Spokesperson. These also provide the voice preview and advanced Spokesperson options used by the Buzz Berd Call plugin. **Speech playback and recovery** - Long Siri replies play to completion through a bounded playback queue. - Session restarts wait for playing and queued speech, including Spokesperson audio, for up to 60 seconds. End Call interrupts immediately. - Failed status sounds release microphone suppression, and native render errors are detected even when the audio engine reports running. - Calls shut down cleanly when their stream reader exits. PCM overflow diagnostics preserve the original error and input ordering. **Siri voices** - Download voices from the menu bar, with support for the macOS 27 download service and earlier versions. - Keep existing voices installed when downloading another, recognize installed assets consistently, and display names for code-named voices. - Preview Apple’s voice samples with `berd-call voices preview`; stopping the preview process stops playback. **Microphone controls** - Automatic input muting persists across session restarts and follows output-device changes, using device names with CoreAudio metadata as a fallback. **Experimental Realtime Spokesperson** - Group the experimental toggle and voice controls above End Call, persist the selected voice, and expose advanced settings through `--spokesperson-options`. - Bring playback loudness close to Siri with gain limiting. - Abort voice changes promptly, recover interrupted input resets, and acknowledge pending input controls in order before shutdown. - Allow separate Realtime, transcription and synthesis keys through `OPENAI_REALTIME_API_KEY`, `OPENAI_STT_API_KEY` and `OPENAI_TTS_API_KEY`, with `OPENAI_API_KEY` as the shared fallback. ### Related issue Related Berd Call work: [persistent Auto input muting](#403), [long Siri replies](#405) and [separate OpenAI service keys](#415). ### Testing - Native Siri downloads preserve previously installed voices; voice samples play through the default output and stop on SIGTERM. A Siri reply longer than a minute plays to completion. - Native call checks verify automatic muting across restarts and output-device changes. Deterministic regressions verify that restarts wait for audio ingress and playback drain. - Local Realtime integration verifies voice-change cancellation, reset ordering and shutdown acknowledgements. The production CLI audio pipe verifies gain limiting across streamed chunks. - Captured Spokesperson audio measures -19.0 LUFS against Siri’s -18.8 LUFS, with a -0.4 dBTP peak. - Recovery regressions exercise failed status cues and native render errors. Local HTTP integration verifies service-specific key selection and shared-key fallback. --------- Signed-off-by: Nora <49aa1f65411fd096d2e2ec144f1e7aa36fdc76d1b907cfdf7be000c66f9d3b8e@buzz.block.builderlab.xyz> Signed-off-by: John Tennant <jtennant@block.xyz> Signed-off-by: Aaron <f425b86c3b780d43f8d9cf1802429d04dd985bf92eac44fca9468b5ad3887c5e@buzz.block.builderlab.xyz> Signed-off-by: Quinn <a0d9f90f86da5f023099d22b8e3599af162d9423a422d8f385a12815de879cf2@buzz.block.builderlab.xyz> Co-authored-by: Nora <49aa1f65411fd096d2e2ec144f1e7aa36fdc76d1b907cfdf7be000c66f9d3b8e@buzz.block.builderlab.xyz> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Aaron <f425b86c3b780d43f8d9cf1802429d04dd985bf92eac44fca9468b5ad3887c5e@buzz.block.builderlab.xyz> Co-authored-by: Quinn <a0d9f90f86da5f023099d22b8e3599af162d9423a422d8f385a12815de879cf2@buzz.block.builderlab.xyz>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
A Siri reply longer than about a minute of speech failed outright with
Siri synthesis exceeded its bounded PCM queue. Siri renders audio faster than playback drains it, and the native callback refused new audio once 60 seconds were pending, which cancelled the reply.The callback now waits for room when that bound is full. Playback frees room as it goes, so long replies play to the end. The bound still caps memory, and shorter replies never wait on playback. Cancelling a reply releases a waiting callback.
Related issue
None.
Testing
exceeded its bounded PCM queuebefore the change and completed after it.