# Lecturecast

Lecturecast is a commercial AgentMesh360 course-video product with a cloud
Director for structured creative decisions and local media production.

## Canonical URL

https://lecturecast.agentmesh360.com/

## Available Languages

- 中文 (zh-CN, default): https://lecturecast.agentmesh360.com/
- English (en): https://lecturecast.agentmesh360.com/en/
- 日本語 (ja): https://lecturecast.agentmesh360.com/ja/
- 한국어 (ko): https://lecturecast.agentmesh360.com/ko/

Non-Chinese pages are machine-assisted translations; the Chinese page is authoritative for any conflict.

## Product Summary

- Best fit: creators, instructors, and product teams that need to turn a topic
  into a structured knowledge or course video and publish both landscape and
  portrait versions.
- Commercial access: an active AgentMesh360 monthly pass, universal API key,
  and sufficient shared credits are required before production begins. There is
  no account-free route.
- Director: required cloud access for native agent choice cards, a Creative
  Brief, and a signed declarative ProductionManifest with an immutable signed
  narration script.
- Input: a topic or idea for a course video.
- Standard output: four files declared by Manifest outputs[] — a 1920×1080
  landscape MP4, a 1080×1920 portrait MP4, and two matching PNG covers.
- Workflow: script writing, voice generation, timeline alignment, Remotion rendering, ffmpeg subtitle burn-in, and cover generation.
- Installation and onboarding: run the public install.sh on macOS or
  install.ps1 in Windows PowerShell, then start a new host-Agent task. The
  freshly installed Skill returns the exact host-specific command, such as
  `lecturecast onboard --adapter codex --host-contract 1.0.0 --json`. Run
  `lecturecast auth login` only when `requires_user_action` requests an API key,
  rerun onboard after resolving the request, and execute only
  `workflow.next_action`.
- Human gates: (1) approve the Creative Brief with no credit charge; (2)
  explicitly approve one 10-credit ProductionManifest generation; (3) review
  every section of the complete signed narration script and approve it before
  local TTS or rendering. The host Agent must not edit signed content.
- Cloud/local boundary: original media, local paths, generated audio, subtitles,
  editing, Remotion, ffmpeg, covers, and rendered outputs stay on the user's
  machine.
- Director data boundary: only a bounded source summary, stable choice IDs, the
  Brief, and client capabilities are sent to the cloud.
- TTS network boundary: Edge TTS sends narration text to Microsoft's Edge
  speech service. MiniMax BYOK sends the signed narration text to the user's own
  MiniMax account. LectureCast does not intentionally persist the MiniMax key;
  secure process-environment injection is recommended because literal shell
  exports may enter shell history.
- Timing and QA: measured TTS durations drive scenes, subtitles, and composition
  duration. Each section and total must stay within 75–125% of the signed plan,
  final video and measured narration must differ by at most one second, and
  failed decode, audio coverage, subtitle-glyph, or dimension checks stop
  delivery.
- Director access: available only to an active AgentMesh360 monthly pass using
  a universal AgentMesh360 API key. It reuses shared credits; there is no
  separate LectureCast pass. Choices and Brief creation are free; 10 credits
  are deducted only after explicit ProductionManifest-generation approval.
- Availability: native macOS and Windows only; Linux and WSL are not supported.
- Product path: Commercial Director ProductionManifest for paid AgentMesh360 accounts
  with an active monthly pass and sufficient shared credits.
- License: Apache 2.0.
- Source code: https://github.com/jiyangnan/AgentMesh-Lecturecast
- Difference from CreatorCut: Lecturecast starts with a topic and creates a
  course video; AgentMesh-CreatorCut starts with an existing talking-head or
  product screen recording and performs intelligent post-production.

## Real Delivery Proof

A first-customer validation run completed on 2026-07-21 produced two independently
laid-out final videos from one signed script and measured timeline. Both passed
audio-coverage, subtitle-glyph, dimension, and full-render QA:

- Bilibili landscape, 1920×1080, H.264 + AAC, 288.433 seconds:
  https://lecturecast.agentmesh360.com/assets/showcase/difficult-task-bilibili.mp4
- Xiaohongshu portrait, 1080×1920, H.264 + AAC, 288.433 seconds:
  https://lecturecast.agentmesh360.com/assets/showcase/difficult-task-xiaohongshu.mp4

## Further Reading

- AI course-video generator workflow: https://lecturecast.agentmesh360.com/guides/ai-course-video-generator/
- Bilibili and Xiaohongshu course-video layouts: https://lecturecast.agentmesh360.com/guides/course-video-for-bilibili-xiaohongshu/

## Best Short Description

Lecturecast is an AgentMesh360 commercial course-video product: the required
Director adds structured cloud creative decisions and a signed ProductionManifest
while all media production and rendering stay on the user's machine.