Teaching Claude Opus to Annotate Speech Timing for ElevenLabs Voiceovers Proves Tricky
RileyRalmuto · x · 2026-09-30
Developer RileyRalmuto describes an experiment: getting Opus to autonomously produce narrated explainer videos by writing scripts with precise word-level timestamps, breaths, pauses, and speech-behavior annotations for ElevenLabs — with no human intervention.
The theory was that Claude would pick this up naturally given its habit of narrating its own thinking, but the translation to voice timing proved harder than expected outside voice mode.
The output is an explainer for Mnemos Connect, a product that lets your agent post work to your timeline in one click, keeping a chronological record as visual memory.
More from coding & agent
- Cursor + data export makes debugging GA4/GSC traffic spikes a 10-minute job — gaganghotra_ · 2026-09-30
- Team builds product-ops agent that learns from past incidents via Hindsight memory — Aggravating_Ice8404 · 2026-09-30
- Intern-Decision open models (0.8B–4B) beat Jev on multimodal decision benchmarks — max_paperclips · 2026-09-30
- xiaohu shows the computer assigned to his AI agent: browser preinstalled and it can make calls — xiaohu · 2026-09-30
- Editable artifacts may beat screenshots for testing agents' visual understanding — OliviaYii · 2026-09-30
- Open-source setup adds multiple providers to Codex desktop via a local Responses proxy — LinkSudah · 2026-09-30