CLI
Cette page n’est pas encore traduite : voici la version anglaise.
The engine is a Python package (vstudio, in lib/). The Mac app and the Claude Code skill both call these commands; you can run them yourself from a terminal:
cd ~/.claude/skills/video-studio # or your clone of the repositoryexport PYTHONPATH="$PWD/lib:$PYTHONPATH"python3 -m vstudio.intake --helpEvery command that reports something accepts --json for one machine-readable document. Each workflow also has its own scripts under workflows/<name>/scripts; see Workflows.
| Command | What it does |
|---|---|
vstudio.intake |
Say what you want and point at files: inventory, plan, revise, apply. |
vstudio.project |
Every workflow as a recipe; projects, runs, checkpoints, inbox, outputs, calendar. |
vstudio.batch |
Batch production: plan, estimate, run, status, review, package. |
vstudio.cleanup |
Shared speech cleanup: pauses, fillers, repeats, restarts. |
vstudio.export |
One clean master into a file, cover and post stub per platform. |
vstudio.watermark |
Your handle, logo or a generated badge on every export: settings, generate, preview, apply. |
vstudio.reframe |
Face-tracked reframe between aspect ratios. |
vstudio.platform |
Print the platform profiles. |
vstudio.effects |
List and inspect the effect registry. |
vstudio.llm |
AI providers on this machine: list, route, auth, test. |
vstudio.retouch |
Portrait retouch and makeup for photos and frames. |
vstudio.lesson |
Lesson recordings: plan teaching points, sync a camera, render clips, write study notes. |
vstudio.qa |
Interviews and podcasts: speakers, question / answer pairs, render Q&A clips. |
vstudio.bilingual |
Translate captions (bilingual / translated) and write per-language SRT / VTT. |
vstudio.intake
Section titled “vstudio.intake”Say what you want and point at files: inventory, plan, revise, apply.
usage: python -m vstudio.intake [-h] {analyze,plan,revise,apply,goal,goal-edit,goal-more,schema} ...
python -m vstudio.intake analyze | plan | revise | apply | goal | goal-edit | goal-more | schema(every command: --json -> one document). goal: one goal-style request ("make as many videos as youcan from this card") -> a survey of the footage (stories by time + place + light) and a plan ofseveral videos, each with the request its own ``plan`` starts from (goal.py). goal-edit: her editson that plan card (drop / rename / kind / length / merge / platforms). goal-more: more videos fora group already made - new plan cards cut from the clips no video of it plays yet. plan / revise--json-events: one JSON event per line on stdout while it works (logs -> stderr), the progress thedesk's plan card shows: {event: stage | progress, stage: scan | probe | listen | faces |transcribe | model | write, ...} (``inventory`` / ``plan.make_plan`` list the fields), then{event: done, plan} - or {event: error, error} and a non-zero exit.
positional arguments: {analyze,plan,revise,apply,goal,goal-edit,goal-more,schema} analyze inventory files / folders (video, audio, images, pdf / docx / pptx / md / srt) plan natural-language request + materials -> plan JSON revise update a plan from a follow-up instruction apply create the project(s) from a plan (a series when mixed); pilot run optional goal a goal-style request (several videos from a folder / card) -> survey + a plan of videos goal-edit apply plan-card edits to a goal JSON (rebuilds each video's request) goal-more more videos for a made group: new plan cards from the clips no video plays schema the plan JSON Schema
options: -h, --help show this help message and exitvstudio.project
Section titled “vstudio.project”Every workflow as a recipe; projects, runs, checkpoints, inbox, outputs, calendar.
usage: python -m vstudio.project [-h] {recipes,new,plan-items,show,status,preview,context,refresh,run,resume,decisions,autopilot,model,reopen,checkpoint,remake,set,export,list,import,import-copy,import-match,adopt,touch,output,ai,series,inbox,calendar,share,feedback} ...
python -m vstudio.project - every workflow as a recipe; projects of N items (N=1 = one video). See
positional arguments: {recipes,new,plan-items,show,status,preview,context,refresh,run,resume,decisions,autopilot,model,reopen,checkpoint,remake,set,export,list,import,import-copy,import-match,adopt,touch,output,ai,series,inbox,calendar,share,feedback} recipes list recipe manifests new create a project plan-items planner recipes: draft the items (segments of a recording) show manifest + project.yaml + status status items and stages preview preview artifacts context agent hook: compact state + playbook + commands refresh re-read project.yaml + item files, re-plan run run until done or a checkpoint resume run until done or a checkpoint decisions autopilot: the decisions in force (+ --history: every one, with blockers) autopilot autopilot on / off for a project (no run; the next run follows it) model the project's AI: show / switch it (every AI step from the next one on) reopen autopilot: take one decision back - the checkpoint waits for you on the next run checkpoint pending checkpoints / answer one remake redo a standing setting of a finished item (vertical framing, post copy, an explainer's visuals) - only what it changes, no other stage set edit params export collect final files list registered projects (+ adopted work folders, series) import finished videos (+ footage) -> a work folder for second-pass editing import-copy an imported clip's post copy written again for every platform import-match match an imported clip's shots to its footage adopt an existing folder made with the skill -> work record + registry touch register a job folder + report its live status (desk 进行中 lane) output second-pass edit of a finished output (references/OUTPUT_EDIT.md) ai project-level AI edit of every output (grouped per output; needs_rerender) series series presets inbox pending checkpoints across projects calendar publish calendar share static review page for a project / clips (share for review) feedback reviewer feedback: import / list / resolve
options: -h, --help show this help message and exit
python -m vstudio.project - every workflow as a recipe; projects of N items (N=1 = one video). Seereferences/PROJECTS.md.
recipes [--json] [--schema] all recipe manifests + capabilities (--schema: the JSON Schema) new --recipe R --dir P [--name N] [--input key=path ...] [--folder D [--glob G]] [--list F] [--csv F] [--episodes N] [--param k=v ...] [--set JSON] [--series S] [--client C] [--variants JSON] [--auto a,b] [--spec JSON] create a project (items from files / lines / rows / episodes) plan-items --dir P [--replace] [--provider X] [--count N] [--min S --max S] planner recipes: draft the items show | status | preview [--item I] [--stage S] | context [--write] | refresh --dir P [--json] run | resume --dir P [--pilot N] [--confirm-pilot] [--items a,b] [--auto ids|all] [--concurrency k=n] [--autopilot [--spend-cap N] [--no-judge] [--lang en|zh|fr] | --ask-first] [--json | --json-events] run until done or a checkpoint needs you (exit 7); --autopilot: no pilot, every checkpoint decided by the AI judge + rules and recorded; only blockers wait (autopilot.py) autopilot --dir P [--on | --off] [--spend-cap N] [--judge | --no-judge] [--lang L] switch it (no run) decisions --dir P [--history] [--json] autopilot: what it decided, by whom, why model --dir P [--provider X [--model M] | --clear] [--by you|chat|group] [--json] the project's AI (project.yaml ``ai``): every AI step uses it, a running one from its next AI step; a switch goes into the decisions log reopen --dir P --id X [--item I] autopilot: take a decision back (the next run asks you) checkpoint --dir P [--id X] [--item I | --items a,b] [--answer JSON | --answer-file F | --default] [--run] list pending payloads / record an answer remake --dir P --item I --do framing=follow|blur [--do copy] [--json-events] a finished item brought up to date: only the vertical exports re-made / the post copy and every post.md written again (remake.py; older projects too); --do visuals: an explainer's scenes drawn again by the AI, render + export again, the voice kept set --dir P [--item I] (--param k=v ... | --set JSON) edit params in project.yaml, then refresh export --dir P [--out D] [--all] [--items a,b] final files + manifest.json (sha256) list registered projects (lanes) with their state + adopted work folders (kind work) + series adopt DIR [--recipe guess|NAME] [--title T] [--client C] a plain folder made with the skill (final/, REPORT.md, post.md, ...) -> DIR/.vstudio/work.json + registry (works.py) import PATH... [--sources DIR]... [--title T] [--dest D] [--check] [--match] [--json | --json-events] finished videos made anywhere (+ their footage) -> a work folder for second-pass editing (imports.py); --check: only say what is finished / footage; --match: shot-match the footage right away; import-copy --dir D --clip C rewrites a clip's post copy; import-match --dir D [--clip C] [--all] touch DIR [--status running|waiting|done|failed] [--stage S] [--progress 0..1] [--message M] [--eta S] [--needs-you] [--recipe R] [--title T] [--outputs a,b] register a job folder + its live status (heartbeat); a recipe project folder gets only its status (row untouched) output list | show | edit | preview-edl | render | undo | redo | revert | ai | chat | effects | kit --project P --output O 2nd-pass edit of one finished output (references/OUTPUT_EDIT.md): edit --ops JSON | --op NAME --param k=v | --op ai --instruction T [--apply]; render [--quality preview|final] [--targets primary,douyin:vertical|all] [--watermark on|off] a launch-kit film: ai answers look / pacing / feature requests with kit.patch (no ops); kit --set PATCH writes it into launch.config.yaml (then resume re-renders through the kit) ai --project P --instruction T [--outputs all|a,b] [--context JSON] [--timeout 120] [--json | --json-events] project-level AI edit: one model call for every output -> grouped changes per output, or needs_rerender (burned-in text on flattened outputs; answered by rules, no model call) series new --id S --recipe R [--name N] [--set JSON] [--cadence JSON] | show --id S | list | update --id S --set JSON inbox [--json] | inbox answer (--project P --id X [--item I] | --id X / --kind K [--projects a,b]) (--answer JSON | --default) calendar account add --id A --platform P [--times 12:00,19:00] [--per-day N] [--days 0,1,2,3,4] calendar plan --project P [--accounts a,b] [--start ISO] | calendar set --post ID --state S [--at ISO] [--url U] calendar list [--start ISO] [--end ISO] [--account A] | calendar add --account A --at ISO [--title T] share --dir P [--outputs a,b] [--items a,b] [--out D] [--quality small|standard|high] [--no-footer] [--title T] [--expiry-note T] [--reply-to EMAIL] [--owner-name N] [--lang auto|en|zh] [--no-zip] [--scan] static review page (index.html + previews + posters + captions + review.json, + zip); --scan: privacy warnings only feedback import (--file F | --text T | stdin) [--dir P] [--no-pin] | list --dir P [--all] | resolve --dir P --id X reviewer answers -> feedback items (approve -> ready; change -> comment pinned in the clip's chat)--json: one JSON document on stdout; --json-events: one JSON event per line on stdout (logs -> stderr).Exit codes: 0 ok, 1 failed items, 2 refused (budget), 3 paused, 4 pilot waits, 5 bad input, 6 busy, 7 needs you.vstudio.batch
Section titled “vstudio.batch”Batch production: plan, estimate, run, status, review, package.
usage: python -m vstudio.batch [-h] {plan,estimate,run,status,review,job,plan-segments,client,deliver,cleanup-sources,metrics,timing,package,verify-manifest,clean,du,bench,recipes} ...
python -m vstudio.batch - batch video production (Fleet F0). See references/BATCH.md.
positional arguments: {plan,estimate,run,status,review,job,plan-segments,client,deliver,cleanup-sources,metrics,timing,package,verify-manifest,clean,du,bench,recipes} plan spec -> jobs in the batch store estimate time / storage / cost estimate and budget check run run the jobs status job table review HTML review page / apply decisions job one job: show (transcript, edits, captions, QC, exports) | edit | rerun plan-segments transcript -> candidate segments (segments.draft.yaml) client client workspace: init | show | update | list deliver client delivery package (per-platform folders, 文案.md, 排期表, zip) cleanup-sources list (dry run) / delete, with the dry run's code, source files of deliveries past their cleanup date metrics metrics: --batch / --client / --all, JSON or the weekly CSV timing review timing event -> batch store package publish folders + schedule + confirmation code verify-manifest recompute a package confirmation code clean delete regenerable intermediates du disk use bench benchmark table recipes list recipes
options: -h, --help show this help message and exit
python -m vstudio.batch - batch video production (Fleet F0). See references/BATCH.md.
plan SPEC [--batch DIR] [--planner file|claude] [--segments F] [--sync] [--json] spec -> jobs, replan estimate [--batch DIR] [--json] machine / wall time, storage, API cost vs budget run [--batch DIR] [--pilot N] [--confirm-pilot] [--resume] [--retry-failed] [--jobs a,b] [--concurrency asr=1,cpu-render=2] [--json-events] run the stage DAGs (resumable; refuses over budget) status [--batch DIR] [--json] [--events N] terminal table review [--batch DIR] [--apply decisions.json|JSON] [--confirm-kinds a,b] [--accept-policy] [--jobs a,b] [--approve-green] [--json] HTML grid + decisions sheet / ingest (+ bulk answers) job ID [--batch DIR] [--json] [--no-words] one job: transcript + cleanup edits, captions, QC, exports package [--batch DIR] [--per-day N] [--start YYYY-MM-DD] [--times 12:00,19:00] [--json] publish folders verify-manifest MANIFEST recompute a package's confirmation code (JSON) clean [--batch DIR] [--dry-run] delete regenerable intermediates of finished jobs du [--batch DIR] disk use of the batch folder bench [--batch DIR] the benchmark table used by estimate recipes [--json] registered recipes (--json: labels, inputs, row keys)v0.2 (desk; all with --json): plan-segments --source F [--transcript T] [--client C] [--count N] [--min S --max S] [--platforms a,b] [--provider auto|claude|openai|ollama|claude-code|...|none] [--out DIR] transcript -> candidate segments -> segments.draft.yaml client init|show|update|list --client C [--set JSON] client.yaml layered over the persona (effective config) job show ID | job edit --job J --op caption|trim|cut|notes|hook|cover|copy|undo ... | job rerun --job J [--json-events] deliver [--client C] [--zip] [--cleanup-days N] [--out DIR] client delivery package + manifest hash cleanup-sources [--batch B | --client C | --all] [--confirm-delete CODE] list (dry run + code) / delete exactly that list; sources outside the batch folder: report only metrics --batch B | --client C | --all [--csv] metrics JSON / weekly CSV timing --job J --event start|stop|add --what review [--seconds S] review timing -> batch store--json: one JSON document on stdout (paths absolute); run --json-events: one JSON event per line on stdout(the human log goes to stderr). See references/BATCH.md section 7.--batch defaults to the current folder when it holds batch.db.Exit codes (run): 0 ok, 1 some jobs failed, 2 refused over budget, 3 paused (circuit breaker), 4 pilot waits.vstudio.cleanup
Section titled “vstudio.cleanup”Shared speech cleanup: pauses, fillers, repeats, restarts.
usage: python -m vstudio.cleanup [-h] {analyze,review,apply,verify} ...
Speech cleanup, one shared tool: 气口 (pauses / breaths), 口头禅 fillers, repeats, stammers,
positional arguments: {analyze,review,apply,verify} analyze detect edits -> cleanup.json + cleanup_review.md review print the review sheet of an EDL apply cut the source with the decisions -> versioned output verify re-ASR the output and flag lost content words (exit 1)
options: -h, --help show this help message and exitvstudio.export
Section titled “vstudio.export”One clean master into a file, cover and post stub per platform.
usage: python -m vstudio.export [-h] [--platforms PLATFORMS] [--out OUT] [--cues CUES] [--captions {both,burn,sidecar}] [--cover COVER] [--cover-canvas {platform,video}] [--cover-ab COVER_AB] [--no-captions] [--title TITLE] [--post POST] [--lang LANG] [--bilingual] [--account ACCOUNT] [--mode {face,subject,center,pad-blur,letterbox,follow,blur}] [--shots SHOTS] [--fallback {subject,center,pad-blur,letterbox}] [--start START] [--dur DUR] [--encoder ENCODER] [--preset PRESET] [--watermark {auto,on,off}] master
Multi-platform export: one clean master -> a file + cover + post stub per platform, plus amanifest.
positional arguments: master clean (caption-free) master video
options: -h, --help show this help message and exit --platforms PLATFORMS comma list of name[:orientation], e.g. xiaohongshu:vertical,douyin,youtube (default persona platforms.default). Known: xiaohongshu:vertical, xiaohongshu:full, xiaohongshu:horizontal, douyin:vertical, douyin:horizontal, tiktok:vertical, youtube:horizontal, youtube- shorts:vertical, bilibili:horizontal, bilibili:vertical, wechat- channels:vertical, wechat-channels:horizontal, x:horizontal, x:square, x:vertical, instagram:reels, instagram:feed, facebook:reels, facebook:feed, facebook:horizontal, linkedin:horizontal, linkedin:square, linkedin:feed, threads:vertical, threads:horizontal, reddit:horizontal, reddit:square, reddit:vertical, pinterest:vertical, pinterest:feed, pinterest:square, snapchat:vertical, kuaishou:vertical, kuaishou:horizontal, weibo:horizontal, weibo:vertical, zhihu:horizontal, zhihu:vertical, dailymotion:horizontal, dailymotion:vertical, kwai:vertical --out OUT --cues CUES cues.json (subs.Cue dicts) or .srt to burn per platform; repeatable as lang=path, one track per language (--cues en=cues.en.json --cues zh=cues.zh.json): each platform takes the track in its language (YouTube English, 小红书 / B站 / 抖音 Chinese) --captions {both,burn,sidecar} both (default): burn + a <target>.<lang>.srt where the platform takes a caption upload (YouTube, X, Facebook, LinkedIn); sidecar: no burn there (clean video + SRT), still burned elsewhere; burn: burn only, no SRT --cover COVER cover image(s): a path (closest aspect wins; another aspect is fitted on a blurred pad) or platform[:orientation]=path for one target, e.g. --cover xiaohongshu=c34.png --cover-canvas {platform,video} cover on the platform canvas or on the exported video canvas --cover-ab COVER_AB thumbnail A/B pack for one target: platform=a.jpg,b.jpg,c.jpg (YouTube Test & Compare) -> <target>.cover.ab1..3.jpg + a note in its post --no-captions do not burn --cues (the master already carries its captions) --title TITLE post title (checked against each platform's limit) --post POST post.json {title, hook, body, chapters, links, tags} -> <target>.post.md --lang LANG post copy language (en | zh); default: the content language detected from the cues (English content -> English copy on X / Instagram / TikTok / YouTube) --bilingual post copy in English + Chinese (post.json en + zh blocks) --account ACCOUNT account tier for limits, e.g. premium (X: longer videos/posts) --mode {face,subject,center,pad-blur,letterbox,follow,blur} reframe mode when the aspect changes: face (talking head), subject / follow (travel, drone: follow what stands out per shot), pad-blur / blur, center, letterbox --shots SHOTS shots.json {transitions: [[t0, t1], ...]} of the master (crossfades) for --mode subject --fallback {subject,center,pad-blur,letterbox} --start START --dur DUR --encoder ENCODER H.264 encoder (default $VSTUDIO_H264_ENCODER / persona export.h264_encoder / libx264): libx264 | videotoolbox | h264_videotoolbox | h264_mf --preset PRESET --watermark {auto,on,off} the creator's watermark (python -m vstudio.watermark): auto = her default, on, offvstudio.watermark
Section titled “vstudio.watermark”Your handle, logo or a generated badge on every export: settings, generate, preview, apply.
usage: python -m vstudio.watermark [-h] {show,set,generate,preview,apply} ...
Watermark: your handle, your logo or a generated badge on every export (off until you set one up).
positional arguments: {show,set,generate,preview,apply} show the effective settings and whether exports get the mark set change the settings (written to $VSTUDIO_HOME/watermark.json) generate draw a logo-style watermark PNG from a handle preview the mark over a sample frame apply re-encode a finished video with the mark
options: -h, --help show this help message and exitvstudio.reframe
Section titled “vstudio.reframe”Face-tracked reframe between aspect ratios.
usage: python -m vstudio.reframe [-h] [--size SIZE] [--platform PLATFORM] [--mode {face,subject,center,pad-blur,letterbox}] [--fallback {subject,center,pad-blur,letterbox}] [--start START] [--dur DUR] [--zoom ZOOM] [--every EVERY] src dst
Reframe a video to another canvas: face-tracked virtual camera, centre crop, blurred-fill orletterbox.
positional arguments: src dst
options: -h, --help show this help message and exit --size SIZE WxH (default from --platform) --platform PLATFORM e.g. douyin or xiaohongshu:vertical (size + safe box) --mode {face,subject,center,pad-blur,letterbox} --fallback {subject,center,pad-blur,letterbox} --start START --dur DUR --zoom ZOOM --every EVERY detect every N framesvstudio.platform
Section titled “vstudio.platform”Print the platform profiles.
youtube:horizontal 1920x1080 safe=(96, 54, 1824, 1026) caption=(96, 880, 1824, 1020) cover=(1280, 720) title<=100 len [420, 1200]/43200s -14.0 LUFS/-1.5 dBTPyoutube-shorts:vertical 1080x1920 safe=(60, 180, 960, 1530) caption=(120, 1260, 960, 1480) cover=(1080, 1920) title<=100 len [20, 60]/180s -14.0 LUFS/-1.5 dBTPtiktok:vertical 1080x1920 safe=(60, 130, 940, 1436) caption=(140, 1200, 940, 1420) cover=(1080, 1920) title<=55 len [21, 60]/3600s -14.0 LUFS/-1.5 dBTPinstagram:reels 1080x1920 safe=(60, 285, 950, 1470) caption=(170, 1200, 910, 1440) cover=(1080, 1920) title<=0 len [15, 90]/1200s -14.0 LUFS/-1.5 dBTPinstagram:feed 1080x1350 safe=(60, 60, 1020, 1230) caption=(60, 1010, 1020, 1200) cover=(1080, 1350) title<=0 len [15, 90]/1200s -14.0 LUFS/-1.5 dBTPx:horizontal 1920x1080 safe=(96, 54, 1824, 990) caption=(96, 870, 1824, 990) cover=(1920, 1080) title<=0 len [15, 90]/140s -14.0 LUFS/-1.5 dBTPx:square 1080x1080 safe=(60, 54, 1020, 990) caption=(60, 800, 1020, 980) cover=(1080, 1080) title<=0 len [15, 90]/140s -14.0 LUFS/-1.5 dBTPx:vertical 1080x1920 safe=(60, 160, 960, 1540) caption=(120, 1240, 960, 1460) cover=(1080, 1920) title<=0 len [15, 90]/140s -14.0 LUFS/-1.5 dBTPfacebook:reels 1080x1920 safe=(65, 269, 1015, 1248) caption=(65, 1040, 1015, 1240) cover=(1080, 1920) title<=0 len [15, 90]/14400s -14.0 LUFS/-1.5 dBTPfacebook:feed 1080x1350 safe=(60, 60, 1020, 1230) caption=(60, 1010, 1020, 1200) cover=(1080, 1350) title<=0 len [15, 90]/14400s -14.0 LUFS/-1.5 dBTPfacebook:horizontal 1920x1080 safe=(96, 54, 1824, 990) caption=(96, 870, 1824, 990) cover=(1920, 1080) title<=0 len [15, 90]/14400s -14.0 LUFS/-1.5 dBTPlinkedin:horizontal 1920x1080 safe=(96, 54, 1824, 990) caption=(96, 870, 1824, 990) cover=(1920, 1080) title<=0 len [30, 120]/900s -14.0 LUFS/-1.5 dBTPlinkedin:square 1080x1080 safe=(60, 54, 1020, 990) caption=(60, 800, 1020, 980) cover=(1080, 1080) title<=0 len [30, 120]/900s -14.0 LUFS/-1.5 dBTPlinkedin:feed 1080x1350 safe=(60, 60, 1020, 1230) caption=(60, 1010, 1020, 1200) cover=(1080, 1350) title<=0 len [30, 120]/900s -14.0 LUFS/-1.5 dBTPthreads:vertical 1080x1920 safe=(60, 160, 960, 1540) caption=(120, 1240, 960, 1460) cover=(1080, 1920) title<=0 len [15, 90]/300s -14.0 LUFS/-1.5 dBTPthreads:horizontal 1920x1080 safe=(96, 54, 1824, 990) caption=(96, 870, 1824, 990) cover=(1920, 1080) title<=0 len [15, 90]/300s -14.0 LUFS/-1.5 dBTPreddit:horizontal 1920x1080 safe=(96, 54, 1824, 990) caption=(96, 870, 1824, 990) cover=(1920, 1080) title<=300 len [15, 120]/900s -14.0 LUFS/-1.5 dBTPreddit:square 1080x1080 safe=(60, 54, 1020, 990) caption=(60, 800, 1020, 980) cover=(1080, 1080) title<=300 len [15, 120]/900s -14.0 LUFS/-1.5 dBTPreddit:vertical 1080x1920 safe=(60, 160, 1020, 1620) caption=(60, 1300, 1020, 1520) cover=(1080, 1920) title<=300 len [15, 120]/900s -14.0 LUFS/-1.5 dBTPpinterest:vertical 1080x1920 safe=(65, 270, 885, 1130) caption=(195, 920, 885, 1120) cover=(1080, 1920) title<=100 len [6, 60]/900s -14.0 LUFS/-1.5 dBTPpinterest:feed 1000x1500 safe=(60, 60, 940, 1340) caption=(60, 1120, 940, 1330) cover=(1000, 1500) title<=100 len [6, 60]/900s -14.0 LUFS/-1.5 dBTPpinterest:square 1080x1080 safe=(60, 60, 1020, 960) caption=(60, 780, 1020, 950) cover=(1080, 1080) title<=100 len [6, 60]/900s -14.0 LUFS/-1.5 dBTPsnapchat:vertical 1080x1920 safe=(60, 200, 940, 1520) caption=(140, 1220, 940, 1440) cover=(1080, 1920) title<=0 len [10, 60]/60s -14.0 LUFS/-1.5 dBTPxiaohongshu:vertical 1080x1440 safe=(48, 60, 1032, 1290) caption=(120, 1080, 960, 1270) cover=(1080, 1440) title<=20 len [30, 180]/900s -14.0 LUFS/-1.5 dBTPxiaohongshu:full 1080x1920 safe=(60, 240, 1020, 1660) caption=(160, 1420, 920, 1640) cover=(1080, 1920) title<=20 len [30, 180]/900s -14.0 LUFS/-1.5 dBTPxiaohongshu:horizontal 1920x1080 safe=(280, 60, 1640, 1020) caption=(280, 880, 1640, 1020) cover=(1920, 1080) title<=20 len [30, 180]/900s -14.0 LUFS/-1.5 dBTPdouyin:vertical 1080x1920 safe=(60, 160, 930, 1480) caption=(180, 1240, 900, 1460) cover=(1080, 1920) title<=55 len [15, 60]/900s -14.0 LUFS/-1.5 dBTPdouyin:horizontal 1920x1080 safe=(96, 60, 1824, 990) caption=(96, 870, 1824, 990) cover=(1920, 1080) title<=55 len [15, 60]/900s -14.0 LUFS/-1.5 dBTPwechat-channels:vertical 1080x1920 safe=(60, 200, 1020, 1460) caption=(60, 1200, 1020, 1420) cover=(1080, 1440) title<=16 len [15, 120]/28800s -14.0 LUFS/-1.5 dBTPwechat-channels:horizontal 1920x1080 safe=(96, 54, 1824, 1000) caption=(96, 870, 1824, 1000) cover=(1920, 1080) title<=16 len [15, 120]/28800s -14.0 LUFS/-1.5 dBTPbilibili:horizontal 1920x1080 safe=(96, 54, 1824, 1000) caption=(96, 870, 1824, 1000) cover=(1146, 717) title<=80 len [180, 900]/36000s -14.0 LUFS/-1.5 dBTPbilibili:vertical 1080x1920 safe=(60, 200, 930, 1500) caption=(150, 1260, 930, 1480) cover=(1080, 1440) title<=80 len [180, 900]/36000s -14.0 LUFS/-1.5 dBTPkuaishou:vertical 1080x1920 safe=(60, 160, 930, 1520) caption=(180, 1240, 900, 1460) cover=(1080, 1920) title<=0 len [15, 90]/900s -14.0 LUFS/-1.5 dBTPkuaishou:horizontal 1920x1080 safe=(96, 60, 1824, 990) caption=(96, 870, 1824, 990) cover=(1920, 1080) title<=0 len [15, 90]/900s -14.0 LUFS/-1.5 dBTPweibo:horizontal 1920x1080 safe=(96, 54, 1824, 1000) caption=(96, 870, 1824, 1000) cover=(1920, 1080) title<=30 len [30, 300]/900s -14.0 LUFS/-1.5 dBTPweibo:vertical 1080x1920 safe=(60, 180, 960, 1520) caption=(120, 1240, 960, 1460) cover=(1080, 1920) title<=30 len [30, 300]/900s -14.0 LUFS/-1.5 dBTPzhihu:horizontal 1920x1080 safe=(96, 54, 1824, 1000) caption=(96, 870, 1824, 1000) cover=(1920, 1080) title<=30 len [60, 600]/3600s -14.0 LUFS/-1.5 dBTPzhihu:vertical 1080x1920 safe=(60, 160, 960, 1560) caption=(120, 1240, 960, 1460) cover=(1080, 1920) title<=30 len [60, 600]/3600s -14.0 LUFS/-1.5 dBTPdailymotion:horizontal 1920x1080 safe=(96, 54, 1824, 1000) caption=(96, 870, 1824, 1000) cover=(1280, 720) title<=255 len [60, 900]/7200s -14.0 LUFS/-1.5 dBTPdailymotion:vertical 1080x1920 safe=(60, 160, 1020, 1620) caption=(60, 1300, 1020, 1520) cover=(1080, 1920) title<=255 len [60, 900]/7200s -14.0 LUFS/-1.5 dBTPkwai:vertical 1080x1920 safe=(60, 160, 930, 1480) caption=(150, 1240, 930, 1460) cover=(1080, 1920) title<=0 len [15, 60]/300s -14.0 LUFS/-1.5 dBTPvstudio.effects
Section titled “vstudio.effects”List and inspect the effect registry.
usage: python -m vstudio.effects [-h] [--write-md] [--check] [--show ID] [--list] [--engine {pil-frame,ffmpeg,hyperframes,html,audio}] [--category {camera,layout,cards,text,highlight,freeze,transitions,looks,audio,progress,covers}]
Effect registry: every visual / audio effect in video-studio as one declarative catalogue.
options: -h, --help show this help message and exit --write-md regenerate the block in references/EFFECTS.md --check verify every entry point --show ID --list --engine {pil-frame,ffmpeg,hyperframes,html,audio} --category {camera,layout,cards,text,highlight,freeze,transitions,looks,audio,progress,covers}vstudio.llm
Section titled “vstudio.llm”AI providers on this machine: list, route, auth, test.
usage: python -m vstudio.llm [-h] {providers,route,auth,models,test} ...
LLM / ASR / TTS providers on this machine
positional arguments: {providers,route,auth,models,test} providers list LLM, ASR and TTS providers and whether they work here route show which provider each task would use auth auth status | login | logout (see vstudio.llm_auth) models the models each provider can use here (see vstudio.llm_models) test one tiny JSON round-trip
options: -h, --help show this help message and exitvstudio.retouch
Section titled “vstudio.retouch”Portrait retouch and makeup for photos and frames.
usage: python -m vstudio.retouch [-h] [--slim SLIM] [--eye EYE] [--eye-extra EYE_EXTRA] [--shine SHINE] [--shine-feather SHINE_FEATHER] [--smooth SMOOTH] [--light LIGHT] [--makeup MAKEUP] [--body BODY] [--pores PORES] [--tone TONE] [--blemish BLEMISH] [--undereye UNDEREYE] [--neck NECK] [--min-face MIN_FACE] [--natural-cap NATURAL_CAP] [--identity-guard IDENTITY_GUARD] [--lip LIP] [--blush BLUSH] [--brow BROW] [--liner LINER] [--shadow SHADOW] [--contour CONTOUR] [--highlight HIGHLIGHT] [--gloss GLOSS] [--preset {daily,glam,natural,none}] [--lip-shade LIP_SHADE] [--blush-shade BLUSH_SHADE] [--shadow-shade SHADOW_SHADE] [--liner-shade LINER_SHADE] [--glasses {auto,none,thin,thick}] [--faces {main,all}] [--face-strength FACE_STRENGTH] [--no-seg] [--no-region] inp out
Portrait retouch for covers and talking-head frames (stills and video).
positional arguments: inp out
options: -h, --help show this help message and exit --slim SLIM --eye EYE --eye-extra EYE_EXTRA --shine SHINE --shine-feather SHINE_FEATHER --smooth SMOOTH --light LIGHT --makeup MAKEUP --body BODY --pores PORES --tone TONE --blemish BLEMISH --undereye UNDEREYE --neck NECK --min-face MIN_FACE --natural-cap NATURAL_CAP --identity-guard IDENTITY_GUARD --lip LIP --blush BLUSH --brow BROW --liner LINER --shadow SHADOW --contour CONTOUR --highlight HIGHLIGHT --gloss GLOSS --preset {daily,glam,natural,none} --lip-shade LIP_SHADE shade name or #rrggbb --blush-shade BLUSH_SHADE shade name or #rrggbb --shadow-shade SHADOW_SHADE shade name or #rrggbb --liner-shade LINER_SHADE shade name or #rrggbb --glasses {auto,none,thin,thick} --faces {main,all} --face-strength FACE_STRENGTH comma list of per-face strengths, largest face first (e.g. 1,0.6) --no-seg landmark-only skin mask --no-regionvstudio.lesson
Section titled “vstudio.lesson”Lesson recordings: plan teaching points, sync a camera, render clips, write study notes.
usage: python -m vstudio.lesson [-h] {plan,notes,sync,render} ...
Lesson -> knowledge-point clips: segment a class recording by teaching point, plus a recap andstudy notes.
positional arguments: {plan,notes,sync,render} sync offset of the camera file against the main recording (audio / timestamps)
options: -h, --help show this help message and exitvstudio.qa
Section titled “vstudio.qa”Interviews and podcasts: speakers, question / answer pairs, render Q&A clips.
usage: python -m vstudio.qa [-h] {plan,render} ...
Interview / podcast -> Q&A clips: who speaks when, question / answer pairs, answers trimmed tight.
positional arguments: {plan,render}
options: -h, --help show this help message and exitvstudio.bilingual
Section titled “vstudio.bilingual”Translate captions (bilingual / translated) and write per-language SRT / VTT.
usage: python -m vstudio.bilingual [-h] {translate,tracks} ...
Bilingual captions (中英): source-language cues + a translated second line, per-language SRT / VTTtracks.
positional arguments: {translate,tracks} translate cues.json / .srt -> cues with a translated alt line tracks cues (with alt) -> per-language SRT / VTT
options: -h, --help show this help message and exit