Audio Overviews#
Two formats:
- TTS overviews (Gemini 3.1 Flash, single voice
CharonEN /KoreHU, ~1-2 min, MP3 96 kbps) - Deep-dive podcasts (NotebookLM 2-host conversational, ~5-20 min, MP3) — for HN-front-page-worthy entries
Subscribe via RSS feed — new wiki entries and updates are syndicated automatically.
🎙️ Featured — v1.0.10 milestones (2026-05-20)#
| Title | Format | Length | Audio |
|---|---|---|---|
| One dropped column broke fifteen AI tools — v1.0.10 milestones (silent-victim cleanup, B-8 RSI Critic κ=0.708 production-flip, subagent-fanout scale, Sprint-2 gated work) | Deep Dive (NotebookLM 2-host) | ~20 min | EN podcast |
Source audits (NotebookLM input):
- Phase-3 re-extract execution + downstream P0 patch
- Wave-A schema-migration-victim sweep (13 silent victims patched)
- B-8 Critic 100-bullet clean re-sample (κ=0.708)
- B-8 Critic 50-bullet (26 unique) scale-up
- HN-launch refresh — v1.0.10 number-update + final-pick
- Schema-migration downstream-grep checklist
Top-7 HN-worthy entries (with NotebookLM deep-dive podcasts)#
| Wiki | EN TTS overview | HU TTS overview | EN podcast (5 min, 2-host) |
|---|---|---|---|
| Claude Code subagent fanout | EN MP3 | HU MP3 | (pending) |
| mgclient autocommit silent rollback | EN MP3 | HU MP3 | (pending) |
| G-Eval bias mitigation pattern | EN MP3 | HU MP3 | (pending) |
| Memgraph CE feature limits | EN MP3 | HU MP3 | (pending) |
| Memgraph multi-labeling typedness | — | — | EN podcast |
| Reranker cost optimization — not size | — | — | EN podcast |
| Layered eval-cascading pattern | — | — | EN podcast |
Other TTS overviews#
Pipelines#
TTS overview (1-2 min, single voice)#
TL;DR (150-200 words) → Gemini 3.1 Flash TTS preview
↓
base64 PCM L16 24 kHz mono
↓
ffmpeg → MP3 96 kbps
- Voice models:
Charon(EN, neutral male),Kore(HU, natural female accent) - Audio tag:
[warm, friendly, conversational, moderate pace] - Cost: ~$0.001–0.01 per minute (Flash-tier preview pricing)
Deep-dive podcast (5 min, 2-host)#
Wiki .en.md → NotebookLM source (text)
↓
notebooklm generate audio --format deep-dive --length default
↓
2-host conversational MP3 (NotebookLM voices)
- Format:
deep-dive(alternatives:brief,critique,debate) - Length:
default(~5 min);short~2 min,long~10 min - Cost: $0 (Google account)
- Pre-written 2-host scripts (Alex + Jordan) in
.vault-nb/audio-overviews/scripts/
Related#
- Gemini TTS pipeline playbook
- NotebookLM CLI gotchas
- RSS feed — subscribe to wiki updates