Majestic: SIP is not G.711-only any more, and the mic can be processed

[openipc-ai]

An audit of the wiki's Majestic SIP and audio coverage turned up three things that are wrong against merged behaviour, plus a feature documented nowhere. All four are fixed here.

SIP is not G.711-only any more

majestic offers Opus and G.711 in one m=audio line and uses whichever the far end picks. The doorbell how-to said "two-way G.711 audio" in the intro, the call-flow walkthrough, and the /audio.ulaw aside.

The how-to asked readers to break their own camera

This is the one worth reading carefully. Step 3 said:

audio:
  codec: ulaw              # PCMU — also try alaw if your PBX prefers
  srate: 8000

and the 488 Not Acceptable Here entry repeated it ("change audio.codec accordingly").

Neither key has ever been a SIP setting, and following that advice costs you real quality:

• audio.codec selects the codec for RTSP, the /audio.* endpoints and recordings. The SIP stack builds its own SDP offer and never reads it. Setting it to ulaw for the doorbell's benefit only makes everything else on the camera narrowband.
• audio.srate was a genuine constraint once. It is not now: G.711 encoding resamples to 8 kHz from whatever the microphone captures (the fix patched the encoders, not the callers), so a camera at 48 kHz makes correct calls and keeps full-bandwidth recordings.

That second one is exactly the tradeoff a user reported having to make — good recordings or working calls, pick one — and the page was still asking for it after the fix landed.

Both are replaced with a short note explaining why there is no codec or rate line in the SIP config, and the 488 entry now points at the PBX's codec list, which is where the actual mismatch is.

The same stale 8 kHz requirement was repeated for A-law/µ-law RTMP push in majestic-streamer.md and is corrected there too.

VQE was documented nowhere

New in majestic — noise reduction, AGC and a high-pass filter on the microphone capture channel. It gets the config keys, a row in the flavour table, and a section of its own.

Two things about it will generate questions otherwise, so both are stated plainly:

It is Ultimate and SoC-gated, which works out to two of the five chips Ultimate is published for:

chip	SoC code	VQE
Hi3516EV200	3516E200	✅
GK7205V200	7205200	✅
GK7205V500	7205500	—
Hi3516CV200	3518E200	—
Hi3516CV300	3516C300	—

(Majestic also builds it for 3516C500, but no Ultimate image is published for that family, so it does not reach anyone today.)

The keys can exist and the feature still do nothing. The DSP stages are shared libraries the SDK dlopens on demand, and images built before OpenIPC/firmware#2361 do not carry them. The section quotes the log line that says so and points at a firmware update rather than at any majestic setting.

Deliberately not included

The three genuinely undocumented audio.* keys — device, dual, extern — are unconditional in the schema but absent from majestic-config.md, which bills itself as "a reference of what exists". Left for a separate pass so this PR stays about the stale SIP advice and the new feature.

Also noted while auditing, well out of scope here: ru/ has no Majestic documentation at all.

[qodo-free-for-open-source-projects]

PR Summary by Qodo

Correct Majestic SIP audio guidance and document microphone VQE

🐞 Bug fix 📝 Documentation 🕐 20-40 Minutes

AI Description

• Correct SIP guidance to describe Opus/G.711 negotiation without changing global audio settings.
• Remove obsolete 8 kHz requirements from SIP and RTMP instructions.
• Document VQE controls, processing stages, hardware eligibility, and firmware dependencies.

Diagram

graph TD
  CFG["Audio config"] --> MIC["Mic capture"] --> VQE{"VQE enabled?"} -->|Yes| DSP["DSP stages"] --> ENC["Audio encoders"] --> SIP["SIP audio"] --> PBX["Remote PBX"]
  VQE -->|No| ENC
  ENC --> MEDIA["RTSP and media"]

Loading

High-Level Assessment

The current approach is appropriate: correct protocol-specific advice in the doorbell guide, list configuration keys in the reference, and explain VQE behavior and constraints in the streamer guide. A dedicated VQE page was considered but would fragment closely related audio documentation without reducing duplication meaningfully.

Files changed (3) +132 / -14

Bug fix (1) +35 / -12

howto-doorbell-from-camera.mdCorrect SIP codec and sample-rate guidance +35/-12

Correct SIP codec and sample-rate guidance

• Updates the doorbell guide to describe Opus/G.711 negotiation and removes harmful recommendations to force global µ-law and 8 kHz settings. Adds Asterisk Opus guidance and redirects codec mismatch troubleshooting to the PBX codec list.

en/howto-doorbell-from-camera.md

Documentation (2) +97 / -2

majestic-config.mdAdd the VQE configuration reference +15/-0

Add the VQE configuration reference

• Documents the master VQE switch and ANR, AGC, and high-pass filter tuning keys. Notes that availability depends on the Ultimate build and supported SoC.

en/majestic-config.md

majestic-streamer.mdDocument resampling and microphone VQE behavior +82/-2

Document resampling and microphone VQE behavior

• Removes the obsolete 8 kHz RTMP requirement and explains encoder resampling. Adds comprehensive VQE documentation covering capture-wide effects, sample-rate capabilities, supported SoCs, configuration defaults, and missing firmware libraries.

en/majestic-streamer.md

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0) 🎨 UX issues (0) 🔗 Cross-repo conflicts (0) 📜 Skill insights (0)

1. Raw audio cannot autoplay ✓ Resolved 🐞 Bug ≡ Correctness

Description

The guide says opening /audio.ulaw needs nothing installed to play, but this endpoint emits raw
headerless µ-law that requires a player to be given its format, channel count, and capture sample
rate. Users following the prerequisite check in a normal browser may hear nothing and incorrectly
conclude that microphone capture is broken.

Code

en/howto-doorbell-from-camera.md[R26-27]

+  is codec-free, so it needs nothing installed to play), and POST a
+  short raw-audio payload to

Evidence

The repository's audio playback instructions explicitly require the ffmpeg-provided ffplay utility
and invoke /audio.ulaw with -f mulaw, -ar 48000, and -ac 1. They separately identify
/audio.html as the browser-oriented player page, contradicting the added claim that opening the
raw endpoint requires nothing installed.

en/majestic-streamer.md[799-810]
en/howto-doorbell-from-camera.md[23-28]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The microphone verification step claims that `/audio.ulaw` plays without installed tooling, although it is a raw stream requiring explicit playback parameters.
## Issue Context
The streamer guide uses `ffplay` with the µ-law format, sample rate, and channel count. Either direct readers to `/audio.html`, which is the provided player page, or include an appropriate `ffplay` command using the configured `audio.srate`.
## Fix Focus Areas
- en/howto-doorbell-from-camera.md[23-28]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

Tip of the day

💡 Did you know, you can route each action level your way: inline, summary, both, or drop

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗