moda-video-footage

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

moda-video-footage

moda-video-footage

<!-- moda:banner -->
PREREQUISITE — load
moda-core
once per session
(step-0, write contract, free/metered map). Already loaded? Skip ahead. If you cannot load it, the non-negotiables: run
moda doctor --json
before anything;
moda brand list
before creating; writes that pin a revision use your last read's — on
stale_revision
, re-read and retry once (it heals); send the canvas link the moment it exists; stuck or failed?
moda ask "<question>"
— free and fast, never guess.
<!-- /moda:banner -->
<!-- moda:banner -->
前提条件 —— 每个会话加载一次
moda-core
(第0步,签订合约,免费/计量计费映射)。已加载?可直接跳过。若无法加载,必须遵守以下规则:所有操作前先运行
moda doctor --json
;创建前先运行
moda brand list
;固定修订版本的写入操作需使用你最后一次读取的版本 —— 若出现
stale_revision
错误,重新读取并重试一次即可自动修复;画布一旦创建完成立即发送链接;遇到卡顿或失败?使用
moda ask "<问题>"
—— 免费且快速,切勿自行猜测。
<!-- /moda:banner -->

What this lane is

本通道功能

The user has REAL clips on disk — a shoot, a trip, screen recordings — and wants a cut: a film, a reel, a highlight. Analysis stays local (Rushes, the optional toolkit riding
moda rushes
, indexes and searches the footage on this machine and uploads nothing); only the selects that make the cut are uploaded, and the deliverable is always BOTH the editable Moda timeline and the rendered mp4. Generating new footage is moda-video-clip's; motion over designs and type is moda-video-motion's; this lane cuts what already exists.
Read references/footage.md before the first command — it is the whole workflow with its invariants; references/otio.md carries the timeline interchange contract the endgame rides. What follows is the shape, not a substitute for either.
用户磁盘上有真实的视频片段 —— 拍摄素材、旅行记录、屏幕录制 —— 想要进行剪辑:制作成影片、短视频、高光集锦。分析过程完全在本地进行(Rushes是搭载于
moda rushes
的可选工具包,仅在本机对素材进行索引和搜索,不会上传任何内容);只有最终入选的精选素材才会被上传,交付物始终包含可编辑的Moda时间线和渲染后的mp4文件。生成新素材属于moda-video-clip的功能范畴;设计与文字动效属于moda-video-motion的功能范畴;本通道仅用于剪辑已有的素材。
首次执行命令前请阅读 references/footage.md —— 该文档包含完整工作流及固定规则;references/otio.md 包含最终交付所依赖的时间线交换协议。以下内容仅为流程概览,不能替代上述两份文档。

Recipe — folder to film

操作流程 —— 从素材文件夹到成片

  1. Toolkit check:
    moda rushes doctor --json
    . Typed
    rushes_not_installed
    teaches the fix —
    moda rushes install
    (it states what it downloads first; sizes are multi-GiB). Can't install? Degrade honestly: pick selects by eye with your harness's own tools and skip to step 5 — say the selection was manual, never pretend an index ran.
  2. Index:
    moda rushes ingest FOOTAGE --project INDEX
    →
    moda rushes analyze INDEX
    →
    moda rushes status INDEX --json
    . The project directory is persistent memory — keep it, reuse it across revisions, never wipe it to re-ask a question.
  3. Brief + probes: one short exchange for target duration, tone, priorities. Then probe search quality with a few
    moda rushes search INDEX "…" --top-k 8 --json
    queries and
    moda rushes show INDEX mom_… --json
    — look at contact sheets and transcript evidence; captions can hallucinate and rank is not confidence.
  4. Choose:
    moda rushes highlight INDEX --target-minutes N
    for the candidate pool (~2.5× the target, one per duplicate cluster — material, not a film), then
    moda rushes select INDEX mom_… --name … --reason … --json
    in play order, with exact rational trims via
    --items
    when a moment needs shaving. Reason at duplicate-cluster level; consume returned ids.
  5. Extract + upload:
    moda rushes extract INDEX sel_… --handles 2s --out SELECTS --json
    (or upload whole originals when the folder is a few small files),
    moda file upload
    each select, rewrite the media map's
    pending
    entries to
    uploaded
    with the REAL returned asset ids.
  6. Timeline: author
    cut-v001.otio
    and dry-run it first —
    moda edit import CANVAS_REF cut-v001.otio --media-map media-map.json --dry-run
    — against a canvas you created (
    moda canvas create --category animation
    ); then apply the same command with
    --yes
    in place of
    --dry-run
    (replace mode clears the existing timeline, so it refuses unconfirmed). Read the fidelity report; surface every dropped/approximated item. New versions are new files — v002, v003 — never an overwrite of the only cut.
  7. Draft, look, iterate:
    moda export CANVAS_REF --format mp4 --scope main_edit -o draft.mp4
    , sample frames, judge against the brief. Small fixes are
    moda edit apply
    operations against a fresh read; a recut is the next
    .otio
    version.
  8. Deliver: the live canvas link FIRST — the human always gets an editable timeline — then the mp4 and the
    .otio
    version file when the user wants an interchange copy for Resolve/Premiere.
Everything here is unmetered: indexing, uploads, import, export and iteration are free, so iterate on the cut as much as the piece needs. A music bed is an upload away (an audio track on the same timeline); a generated score is moda-audio's lane.
  1. 工具包检查:
    moda rushes doctor --json
    。若返回
    rushes_not_installed
    错误,按照提示执行修复 —— 运行
    moda rushes install
    (命令会先说明需要下载的内容,大小为数GiB)。无法安装?请如实降级处理:使用你当前环境自带的工具人工挑选精选素材,然后直接跳到第5步 —— 请明确说明是人工选择,切勿假装已运行索引。
  2. 索引:依次执行
    moda rushes ingest FOOTAGE --project INDEX
    →
    moda rushes analyze INDEX
    →
    moda rushes status INDEX --json
    。项目目录是持久化存储 —— 请保留该目录,可在不同修订版本中重复使用,切勿为了重新查询而清空目录。
  3. 需求确认 + 搜索测试:简短沟通确认目标时长、风格、优先级。然后通过几个
    moda rushes search INDEX "…" --top-k 8 --json
    查询和
    moda rushes show INDEX mom_… --json
    命令测试搜索质量 —— 查看接触印样和文字记录依据;字幕可能存在幻觉问题,排序结果不代表置信度。
  4. 素材选择:运行
    moda rushes highlight INDEX --target-minutes N
    生成候选素材池(约为目标时长的2.5倍,每个重复素材簇仅保留一个 —— 是素材集合,而非成片),然后按照播放顺序执行
    moda rushes select INDEX mom_… --name … --reason … --json
    选择素材,若需要裁剪某个片段,可通过
    --items
    参数进行精确的时间裁剪。请在重复素材簇层面说明选择理由;使用返回的ID进行后续操作。
  5. 提取 + 上传:运行
    moda rushes extract INDEX sel_… --handles 2s --out SELECTS --json
    提取素材(如果文件夹里只有几个小文件,也可以直接上传原始文件),对每个精选素材执行
    moda file upload
    上传,使用实际返回的资源ID将媒体映射表中的
    pending
    条目更新为
    uploaded
    。
  6. 时间线制作:先创建
    cut-v001.otio
    文件并进行试运行 —— 针对你创建的画布(使用
    moda canvas create --category animation
    创建)执行
    moda edit import CANVAS_REF cut-v001.otio --media-map media-map.json --dry-run
    ;然后将命令中的
    --dry-run
    替换为
    --yes
    正式执行(替换模式会清空现有时间线,因此未确认的操作会被拒绝)。请阅读保真度报告;列出所有被丢弃/近似处理的条目。新版本需保存为新文件 —— 如v002、v003 —— 切勿覆盖唯一的剪辑版本。
  7. 草稿、预览、迭代:运行
    moda export CANVAS_REF --format mp4 --scope main_edit -o draft.mp4
    导出草稿,抽取帧画面,对照需求进行评估。小修小补可以在重新读取时间线后通过
    moda edit apply
    操作完成;重新剪辑则需要生成下一个
    .otio
    版本文件。
  8. 交付:首先发送实时画布链接 —— 用户始终需要可编辑的时间线 —— 然后如果用户需要可在Resolve/Premiere中使用的交换文件,再提供mp4文件和
    .otio
    版本文件。
本通道的所有操作均不计费:索引、上传、导入、导出和迭代都是免费的,因此你可以根据作品需要任意迭代剪辑版本。背景音乐只需上传即可添加(作为同一时间线上的音轨);生成配乐属于moda-audio通道的功能。

Boundaries stated plainly

明确边界

  • Rushes never uploads, never calls cloud inference, and needs no Moda account; Moda's cloud renders and exports and never reads local disk. You are the bridge, and selected media is the only cargo.
  • The user chose uploads of SELECTED media — never bulk-upload the folder.
  • Supported transitions and visual tracks import and render. Unknown effects, unpaired overlaps and resource limits decline by name in
    validation.diagnostics
    ; relay them honestly. Never claim an unsupported edit exported without loss.
  • An audio-only timeline does not export as mp4 — keep a visual clip on the cut.
  • Full inference is Linux x86_64 and Apple Silicon; elsewhere use the media-only profile or the manual degraded path, and say which ran.
  • Rushes 绝不会上传数据,不会调用云端推理,也不需要Moda账户;Moda云端负责渲染和导出,绝不会读取本地磁盘。你是两者之间的桥梁,精选素材是唯一的传输内容。
  • 用户选择仅上传精选素材 —— 切勿批量上传整个文件夹。
  • 支持的转场效果和视频轨道可以正常导入和渲染。未知效果、未配对的重叠内容和资源限制会在
    validation.diagnostics
    中按名称列出;请如实告知用户。切勿声称不支持的编辑操作可以无损导出。
  • 纯音频时间线无法导出为mp4 —— 请在剪辑中保留至少一个视频片段。
  • 完整推理功能仅支持Linux x86_64和Apple Silicon架构;其他架构请使用纯媒体配置或人工降级路径,并明确告知用户使用的是哪种模式。

Errors

错误处理

Any typed error → moda-core's recovery reference.
rushes_not_installed
/
rushes_update_required
: run
moda rushes install
, then re-run the SAME command.
otio_import_rejected
: the fidelity items name each offender — fix them (upload + rewrite
pending
entries; re-reference intermediates with their own timings), never invent asset ids.
stale_revision
on an edit: re-read and retry once — it heals.
任何类型化错误 → 参考moda-core的恢复指南。
rushes_not_installed
/
rushes_update_required
:运行
moda rushes install
,然后重新执行相同命令。
otio_import_rejected
:保真度报告中会列出每个有问题的条目 —— 修复这些问题(上传素材并更新
pending
条目;使用自身时间参数重新引用中间文件),切勿编造资源ID。编辑时出现
stale_revision
:重新读取并重试一次即可自动修复。

Make it recurring

扩展场景

The weekly cut from a growing folder → moda-automate; the finished film's social variants → moda-social; a designed cover still → moda-image.
See also: moda-video — the family fork · moda-video-motion — type and motion over the cut · moda-core — contract, recovery, everything Moda can do.
从不断更新的素材文件夹生成每周剪辑 → 使用moda-automate;为成片制作社交媒体版本 → 使用moda-social;设计封面静图 → 使用moda-image。
另请参阅:moda-video —— 视频系列总览 · moda-video-motion —— 剪辑中的文字与动效 · moda-core —— 合约、恢复机制、Moda的所有功能。

References

参考文档

DocLoad when
references/footage.mdALWAYS before the first rushes command — the workflow, its caching/timing invariants, degraded paths
references/otio.mdbefore authoring or importing a timeline — media maps, fidelity report, exact-time discipline, OTIO vs edit ops
references/export.mdrendering the draft and final — scopes, ceilings, warnings
references/reading-and-verifying.mdthe review loop; verifying without vision
文档加载时机
references/footage.md首次执行rushes命令前必须阅读 —— 包含工作流、缓存/时序规则、降级路径
references/otio.md创建或导入时间线前阅读 —— 包含媒体映射、保真度报告、精确时序规则、OTIO与编辑操作的区别
references/export.md渲染草稿和终稿前阅读 —— 包含作用域、上限、警告事项
references/reading-and-verifying.md评审循环相关;无需视觉能力的验证方法