Repository navigation

AI/ML API as a cloud provider preset (re-ported onto 1.5.6) - #2

Open

Lookoff-AIMLAPI wants to merge 298 commits into

mainfrom

feat/aimlapi-provider-1.5.6

Open

AI/ML API as a cloud provider preset (re-ported onto 1.5.6)#2
Lookoff-AIMLAPI wants to merge 298 commits into
mainfrom
feat/aimlapi-provider-1.5.6

This pull request is big! We're only showing the most recent 250 commits

Commits on Sep 27, 2026
Commits on Sep 28, 2026
- fix: 修复 Studio 处理、监控与 Windows 导入预览 (#221)
  zhouxiaoka
  authored
  ad99613
- docs: 精简 README 产品介绍与入门流程，同步八语文案 (#226)
  zhouxiaoka
  authored
  609e3fb
Commits on Sep 29, 2026
Commits on Sep 30, 2026
- feat: 重排 AI 设置、首次引导弹窗、示例项目与 Studio 编辑器重做
  
  zhouxiaoka
  and
  cursoragent
  committed
  a19cac4
- feat(studio): 按镜头取景——先切镜再分类，无人物镜头改为完整画面
  
  zhouxiaoka
  and
  cursoragent
  committed
  83555a1
- fix(studio): 预览拖到成片范围之前自动吸回起点，脚注标出成片取用区间
  
  zhouxiaoka
  and
  cursoragent
  committed
  65045de
- feat(studio): 预览播放条改为成片自身时间轴（0 到片段时长），不再暴露整段原片进度
  
  zhouxiaoka
  and
  cursoragent
  committed
  2da2399
- docs: CHANGELOG 记录设置重排、首次引导、示例项目与 Studio 取景改动
  
  zhouxiaoka
  and
  cursoragent
  committed
  7aa4e96
- fix: 示例项目原片 source.mp4 加入版本库（原被 *.mp4 忽略，CI 与打包都缺文件）
  
  zhouxiaoka
  and
  cursoragent
  committed
  af71893
- 设置页重排 + 首次引导 + 示例项目 + Studio 按镜头取景 (#232)
  zhouxiaoka
  authored
  c262575
- feat(analytics): connect onboarding and Studio delivery monitoring (#233)
  zhouxiaoka
  authored
  e2b2e49
- feat: centralize platform output strategies
  
  zhouxiaoka
  and
  claude
  committed
  ea66c32
- feat(studio): add platform generation contract
  
  zhouxiaoka
  and
  claude
  committed
  c689954
- feat(studio): auto-generate platform variants
  
  zhouxiaoka
  and
  claude
  committed
  6b196f7
- feat(studio): make platform output the default flow
  
  zhouxiaoka
  and
  claude
  committed
  a77cbdc
- feat(render): append optional AutoClip outro
  
  zhouxiaoka
  and
  claude
  committed
  9d53fa6
- feat(studio): append and retry output variants
  
  zhouxiaoka
  and
  claude
  committed
  3bcc736
- Fix Whisper recovery, processing settings and bounded video imports (#236)
  zhouxiaoka
  authored
  48a9435
- feat(analytics): observe quick output outcomes
  
  zhouxiaoka
  and
  claude
  committed
  b213975
- feat: add 88API sponsor provider and setup guides (#235)
  zhouxiaoka
  authored
  eaef5b0
- fix: recover timeline candidates and scoring, align project cards (#237)
  zhouxiaoka
  authored
  0c84b11
- feat: verify 1080p60 exports and installed Windows video pipeline (#239)
  zhouxiaoka
  authored
  5bae596
- feat(studio): preview completed output variants
  
  zhouxiaoka
  and
  claude
  committed
  1dc9532
- fix: recover compatible timelines and report actual failure causes (#240)
  zhouxiaoka
  authored
  4addf2a
- feat(publish): route publishing through completed output variants
  
  zhouxiaoka
  and
  claude
  committed
  06c945b
- fix: defer blocked SQLite WAL setup without blocking project reads (#241)
  zhouxiaoka
  authored
  4d040cd
- fix: unblock desktop packaging of on-demand Whisper dependencies (#242)
  zhouxiaoka
  authored
  3575831
- fix(studio): render every automatic variant with bounded CPU
  
  zhouxiaoka
  and
  claude
  committed
  25cb051
- feat(i18n): translate fast-output copy and platform names
  
  zhouxiaoka
  and
  claude
  committed
  d31425f
- fix(branding): keep the outro visible for its full second
  
  zhouxiaoka
  and
  claude
  committed
  757bc1b
- fix(platforms): enforce only real platform length limits
  
  zhouxiaoka
  and
  claude
  committed
  2c56a85
- feat(studio): skip our subtitles when the source already has them
  
  zhouxiaoka
  and
  claude
  committed
  84f0877
- feat(studio): copy share caption and ask for a rating after download
  
  zhouxiaoka
  and
  claude
  committed
  3b07f0a
- fix(studio): keep rendering state alive and lower encode threads
  
  zhouxiaoka
  and
  claude
  committed
  fda1a57
- fix(render): dim and cheaply blur the vertical backdrop
  
  zhouxiaoka
  and
  claude
  committed
  5737a6d
- feat(studio): frame automatic vertical outputs on the speaker
  
  zhouxiaoka
  and
  claude
  committed
  ec42ba3
- fix(studio): accept sources longer than two hours
  
  zhouxiaoka
  and
  claude
  committed
  52eaba5
- feat: add SenseVoiceSmall as a local subtitle transcription option (#244)
  zhouxiaoka
  authored
  e9e6c90
- refactor: 整理旧运维入口与过时仓库文档 (#245)
  zhouxiaoka
  authored
  7fd70c9
- docs: refresh contribution guide and GitHub community entrypoints (#246)
  zhouxiaoka
  authored
  70e4f42
- fix(studio): bound ffmpeg CPU in framing, preview and legacy export
  
  zhouxiaoka
  and
  claude
  committed
  4941753
- feat(cli): 支持 opencode MCP 一键接入（autoclip mcp install opencode） (#229)
  
  namana843-bit
  and
  zhouxiaoka
  authored
  8968916
- fix(import): ship yt-dlp's YouTube challenge solver and use any JS runtime
  
  zhouxiaoka
  and
  claude
  committed
  bdc74e3
- feat(studio): package automatic vertical outputs with platform templates
  
  zhouxiaoka
  and
  claude
  committed
  ab79068
- feat(studio): show template packaging and allow title edits
  
  zhouxiaoka
  and
  claude
  committed
  c105acc
- feat(studio): two-line caption screens and template styles
  
  zhouxiaoka
  and
  claude
  committed
  910d1c8
- fix(packaging): translate sentence segments instead of subtitle rows
  
  zhouxiaoka
  and
  claude
  committed
  ed2605c
- Merge origin/main into worktree-project-diagnosis
  
  zhouxiaoka
  and
  claude
  committed
  f006b08
- feat(analytics): observe automatic generation and template packaging
  
  zhouxiaoka
  and
  claude
  committed
  506beb7
- fix(packaging): keep titles and bilingual captions inside the frame
  
  zhouxiaoka
  and
  claude
  committed
  853a13c
- fix(studio): end clips on complete sentences and thoughts; caption burned foreign sources
  
  zhouxiaoka
  and
  claude
  committed
  d2f97bc
- fix(studio): cut inside rows at sentence ends and never end on the next question
  
  zhouxiaoka
  and
  claude
  committed
  2272d84
- fix(studio): end clips in a real pause after a finished passage
  
  zhouxiaoka
  and
  claude
  committed
  206d742
- feat(packaging): pick palette and style from the content mood
  
  zhouxiaoka
  and
  claude
  committed
  a7b6b57
- docs: record boundary snapping and mood looks
  
  zhouxiaoka
  and
  claude
  committed
  088dde8
- fix(packaging): retry rejected responses, plain audience-language titles, varied batch palettes
  
  zhouxiaoka
  and
  claude
  committed
  ecad8d9
- fix(packaging): clean fallback titles taken from the draft hook
  
  zhouxiaoka
  and
  claude
  committed
  1931c51
Commits on Oct 1, 2026
- perf(studio): trim the fast-output chain and record model token usage
  
  zhouxiaoka
  and
  claude
  committed
  628b01b
- docs: fix the fast-output workflow and what was trimmed
  
  zhouxiaoka
  and
  claude
  committed
  822886e
- perf(pipeline): run per-chunk model calls side by side
  
  zhouxiaoka
  and
  claude
  committed
  226b3db
- feat(pipeline): find clips in one pass over the whole transcript
  
  zhouxiaoka
  and
  claude
  committed
  5aa5f0f
- perf(render): encode studio clips with the hardware H.264 encoder
  
  zhouxiaoka
  and
  claude
  committed
  5946cd7
- feat(studio): render the top 10 clips automatically, the rest on demand
  
  zhouxiaoka
  and
  claude
  committed
  f02be3e
- perf(studio): use the creator's uploaded subtitles instead of speech recognition
  
  zhouxiaoka
  and
  claude
  committed
  ceb735e
- perf(studio): package a batch of clips side by side
  
  zhouxiaoka
  and
  claude
  committed
  6cfe7e4
- docs: record pipeline V2 progress and regression numbers
  
  zhouxiaoka
  and
  claude
  committed
  355ee9b
- docs: rewrite README around real outputs and V2 speed/cost numbers
  
  zhouxiaoka
  and
  cursoragent
  committed
  0ee5633
- feat(studio): every output ships a publish kit: cover, post copy, bundle
  
  zhouxiaoka
  and
  claude
  committed
  0123851
- feat(publish): publish an output with its kit copy and tags
  
  zhouxiaoka
  and
  claude
  committed
  4ca3ff4
- docs: measured cost and time per video link
  
  zhouxiaoka
  and
  claude
  committed
  05b33bf
- docs: point README to the case library and add a showcase submission template
  
  zhouxiaoka
  and
  cursoragent
  committed
  88189f2
- fix(studio): re-download YouTube sources that arrive at 360p
  
  zhouxiaoka
  and
  claude
  committed
  09a5ffa
- feat(branding): the designed 1.8 s outro animation replaces the text card
  
  zhouxiaoka
  and
  claude
  committed
  1dc8a09
- fix(packaging): caption burned-in sources only for audiences that cannot read them
  
  zhouxiaoka
  and
  claude
  committed
  5c91702
- docs: README leads with one link, one click; adds the TIM run and stage timing
  
  zhouxiaoka
  and
  cursoragent
  committed
  0c2b275
- feat(studio): output cards carry their publish kit
  
  zhouxiaoka
  and
  claude
  committed
  35ece30
- test(bench): fast-output regression set with per-stage timing
  
  zhouxiaoka
  and
  claude
  committed
  53ececd
- docs(design): serif web headlines and the risograph art layer
  
  zhouxiaoka
  and
  cursoragent
  committed
  d2cf651
- docs: lead README with 10+ clips per link; drop competitor names and roadmap hints
  
  zhouxiaoka
  and
  cursoragent
  committed
  7fd9873
- perf(asr): upload cloud transcription chunks side by side
  
  zhouxiaoka
  and
  claude
  committed
  6786f2e
- feat(settings): dedicated Bailian endpoints for DashScope connections
  
  zhouxiaoka
  and
  claude
  committed
  a306660
- perf(asr): 8 parallel cloud transcription uploads; benchmark keeps case order
  
  zhouxiaoka
  and
  claude
  committed
  82f82e0
- fix(cover): long English titles rewrap larger and push the photo down
  
  zhouxiaoka
  and
  claude
  committed
  882a59d
- docs: tighten README layout so the first screen and proof stay scannable
  
  zhouxiaoka
  and
  cursoragent
  committed
  3a69c63
- feat(cover): the vision model picks the guest's frame; AI covers upgrade the design
  
  zhouxiaoka
  and
  claude
  committed
  47b6f79
- docs: sync the seven README translations to the one-link, one-click rewrite
  
  zhouxiaoka
  and
  cursoragent
  committed
  95153d1
- fix(packaging): one language per platform; no false burned captions
  
  zhouxiaoka
  and
  claude
  committed
  e7fb0c1
- feat(cover): fal image connections (GPT Image 2.5) and correct-by-construction text
  
  zhouxiaoka
  and
  claude
  committed
  41fa605
- fix(settings): an unknown image API never invalidates the settings file
  
  zhouxiaoka
  and
  claude
  committed
  4d11ece
- feat(cover): designed covers by default; AI covers only when the user picks a model
  
  zhouxiaoka
  and
  claude
  committed
  7814dbd
- docs: 1.5.0 fast-output handoff, kept current as work lands
  
  zhouxiaoka
  and
  claude
  committed
  ed14a34
- docs(handoff): demos rebuilt; note 9:16 AI cover padding
  
  zhouxiaoka
  and
  claude
  committed
  1317b5d
- fix(cover): AI covers fill the platform frame, no flat bars
  
  zhouxiaoka
  and
  claude
  committed
  0fc14a6
- fix(studio): show the whole cover in the kit; drop dead raw-clip cards
  
  zhouxiaoka
  and
  claude
  committed
  a0108d8
- fix(studio): backups render only once prepared; review fixes
  
  zhouxiaoka
  and
  claude
  committed
  3b8088c
- fix(studio): English platforms stay English on every path
  
  zhouxiaoka
  and
  claude
  committed
  6931bc1
- fix: failures degrade instead of losing work; record the review
  
  zhouxiaoka
  and
  claude
  committed
  ecd0dda
- feat(studio): blur burned Chinese captions out of English versions; 1.4 AI covers off
  
  zhouxiaoka
  and
  claude
  committed
  723eefb
- fix(1.5): complete output acceptance and align CLI/MCP
  zhouxiaoka
  committed
  d3e3cb1
- fix(ci): isolate runtime tests and sync quick-output docs
  zhouxiaoka
  committed
  5c0b82e
- fix(cli): apply Windows connection cleanup at headless startup
  zhouxiaoka
  committed
  e23dc99
- fix(release): bundle face detector and verify installed portrait output
  zhouxiaoka
  committed
  e3e5723
- fix(render): select bundled CJK family across font providers
  zhouxiaoka
  committed
  ee73e59
- docs(1.5): record cross-platform installed wheel acceptance
  zhouxiaoka
  committed
  1d433ea
- fix: complete 1.5 delivery telemetry and publish CLI MCP assets
  zhouxiaoka
  committed
  0147056
- fix: settle observed AI cover results before replacement
  zhouxiaoka
  committed
  0fe48c0
- fix(studio): rank eligible clips per platform and settle empty queues
  zhouxiaoka
  committed
  1b02347
- Merge pull request #250 from zhouxiaoka/worktree-project-diagnosis
  zhouxiaoka
  authored
  e940f5a
- docs(1.5): record merged tag and final long-video acceptance
  zhouxiaoka
  committed
  9a130ee
- Merge pull request #251 from zhouxiaoka/worktree-project-diagnosis
  zhouxiaoka
  authored
  9895842
- docs: record 1.5.0 latest release and issue closure
  zhouxiaoka
  committed
  52b20fa
- Merge pull request #252 from zhouxiaoka/worktree-project-diagnosis
  zhouxiaoka
  authored
  d3e26ee
- docs: align eight READMEs with the 1.5.0 release
  zhouxiaoka
  committed
  8f9416f
- docs: review current behavior and remove stale README claims
  zhouxiaoka
  committed
  7d33f3c
- docs: preserve the existing sponsor sections verbatim
  zhouxiaoka
  committed
  ce35f98
- Merge pull request #248 from zhouxiaoka/cursor/readme-v2
  zhouxiaoka
  authored
  c83fca0
- docs: add sixteen playable video demos to translated READMEs
  zhouxiaoka
  committed
  a8cac39
- docs: align video thumbnails across cards
  zhouxiaoka
  committed
  51c2f09
- Merge pull request #253 from zhouxiaoka/codex/readme-video-demos
  zhouxiaoka
  authored
  e22be4b
Commits on Oct 2, 2026
Commits on Oct 3, 2026
Commits on Oct 4, 2026
Commits on Oct 5, 2026
Commits on Oct 6, 2026
- Allow a named owner waiver to ship 1.5.4 (#292)
  
  zhouxiaoka
  and
  cursoragent
  authored
  cf6fad4
- chore: sync main with the released 1.5.5 version and changelog (#293)
  
  zhouxiaoka
  and
  cursoragent
  authored
  8538e15
- docs: put the official website first in the eight READMEs (#255)
  
  zhouxiaoka
  and
  cursoragent
  authored
  b77c930
- fix(studio): fall back to the subtitle route when optional auto screening fails
  
  zhouxiaoka
  and
  cursoragent
  committed
  576e0b4
- Ship 1.5.6: ready imports before start, classify failures, and drop desktop Celery
  
  zhouxiaoka
  and
  cursoragent
  committed
  347544c
- Ship 1.5.6: ready imports, failure taxonomy, desktop without Celery (#295)
  zhouxiaoka
  authored
  f05ef42
Commits on Oct 7, 2026
- feat: add AI/ML API as a cloud provider preset
  Lookoff-AIMLAPI
  committed
  be4bed3

Commits

Commits on Sep 27, 2026

Commits on Sep 28, 2026

Commits on Sep 29, 2026

Commits on Sep 30, 2026

Commits on Oct 1, 2026