Skip to content

v0.28.0

Choose a tag to compare

@github-actions github-actions released this 12 Nov 18:33
· 74 commits to main since this release
9b0c90d

Highlights

3D in the toolkit for the first time, terminal demos, and publishing in one step. The toolkit's own video, Made with itself, ships as an example.

  • Blender shot specs (experimental, optional): blender_render.py turns a JSON spec into a 3D film, one clip per shot for any build, or transparent clips that BlenderClip plays inside a Remotion video. A rigged cartoon cast is described in JSON, acting lands on narration words, and check verifies every camera by geometry before anything renders. Needs Blender 4.2+; without it nothing changes (#214).
  • Terminal demos: the Terminal component types commands and prints output on narration cues, and brands/video-toolkit gives the toolkit its own look (#207).
  • Several narrators per brand: pick one per video with voiceover.py --voice (#207).
  • Publish kit: publish_kit.py makes one whole-video SRT and chapters from the render, /publish passes both to the upload, and youtube_upload.py --check says what a login can do and how to fix what it can't (#211, #209).
  • Show and tell: share what you make in Discussions; /contribute drafts the post for you (#216).

Thanks to @jdela141252-lab for two fixes: modal_setup.py deploys on a newly created Modal workspace (#212), and qc.py --layout works on Windows (#213).

What's Changed

  • Release v0.28.0: Blender shot specs (experimental), the Terminal component, several narrators per brand, publish kit and youtube_upload --check, Made with itself as an example, Show and tell (bb200e2)
  • DOCS: phones get the Release 650 picture as the site banner, since the Made with itself terminal frame can't be read at that width (e093729)
  • FIX: copy_lint skips capture/ folders, saved command output kept verbatim for the screen, not copy anyone wrote (a47fccb)
  • DOCS: the README and site lead with Made with itself, add the design-sources video, link the parody videos on YouTube, and mark the agent handoff as an experiment (14542c7)
  • FEAT: examples/toolkit-made-itself, the toolkit's own 93s video made in the toolkit: every terminal line from capture/ (pick() fails the bundle if one is missing), the bill measured, the QC finding real (19acaba)
  • FIX: BarChart grows a highlighted value by scaling it instead of changing its font size every frame, which crashed Chrome mid-render on Remotion 4.0.539 ("Page crashed!") (014e8ef)
  • FIX: Show and tell keeps the showcase table whole, and /contribute option 6 checks the link is not private, finds the title and render when a template leaves them out, and passes the post title as text (cf52b16)
  • FEAT: /contribute can share a video you made in Show and tell: it drafts the post from project.json and the spend log, lints it and posts only after a yes, and /publish offers it after an upload (6b8d605)
  • DOCS: the README points people to Show and tell in Discussions to share the videos they make (43508db)
  • FIX: a white picture on a Blender screen (a document or a light UI on a laptop, phone or card) no longer fails check as blown out (b6b16e3)
  • FIX: the Blender product-shot example passes its checks under Cycles as well as EEVEE: its backdrop is a step darker, since Cycles measures about 0.01 brighter and one shot sat on the exposure limit (0803d60)
  • DOCS: /video suggests a Blender shot when Blender is installed, /setup and /versions report it as optional, and the roadmap has the next round (7361f98)
  • FEAT: three Blender examples: a character film from one spec, a product stage with a phone, a card and a 3D model, and Blender clips inside a Remotion video (5b99d27)
  • FEAT: BlenderClip plays Blender renders in a Remotion video: one clip per shot, transparent if asked, with the shot's own sound and the frame of each acting event (55cf341)
  • FEAT: Blender shot specs (experimental, optional): blender_render.py turns a JSON spec into 3D footage, with a rigged cartoon cast described in JSON, a desk set and a product stage, acting on narration cues, and every camera checked by geometry before a render (d2d0020)
  • FIX: qc.py reads subprocess output as UTF-8, so --layout no longer crashes on Windows when on-screen text has curly quotes, and an empty or broken audit result is reported instead of raising a TypeError (5a18845)
  • FIX: modal_setup.py deploys on a new Modal workspace: a refused image builder version (2026.10) is retried with 2025.06, which the rest of the run keeps (d99bb6f)
  • FEAT: tools/publish_kit.py makes one whole-video SRT and a chapter at each scene's cut from a single transcription of the render, and /publish checks the setup, runs the kit and passes both to the upload (3a76ca5)
  • FEAT: youtube_upload.py --check says which channel a login is for, which scopes it has and whether the Data and Analytics APIs are on, with the fix for each, and youtube_stats links the page that switches Analytics on (f7a1edb)
  • DOCS: /video offers music after the voiceover: music_gen takes, pick_music's measured pick, then audition.py renders for the user to choose by ear (036066f)
  • FIX: contact_sheet.py takes --at more than once (it kept only the last), and music_gen.py and contact_sheet.py lose a stray f-string and an unused import (d65ad3d)
  • FIX: video jobs draw a local background one step off the caption pill's ink, so the pill shows on it instead of vanishing into the same colour (e804f5e)
  • FIX: qc.py messages use colons and commas instead of dashes, so a finding quoted on screen or in copy passes the copy lint (8f1d524)
  • FIX: youtube_upload asks for the captions scope, refuses --captions on an older token before uploading, and adds captions, a thumbnail or a playlist to an uploaded video with --video-id (70962d1)
  • FIX: Terminal draws a line on the frame it is cued for, so a still at that frame (a thumbnail) shows it (0bdf384)
  • FEAT: a brand can hold several narrators, picked per video with voiceover.py --voice, and brand ElevenLabs model, settings and maxWpm now apply; the toolkit brand narrates with Marilyn and Bren (27f0e7b)
  • FEAT: lib/components Terminal types commands and prints output on narration cues, with menus and scrolling, and brands/video-toolkit gives the toolkit its own look from the site's tokens (c629d56)
  • DOCS: the roadmap adds a tutorial template with batch uploads, a Terminal component, translating a finished video and a publish kit, and ticks transcript highlights (done in #162) (1f8fd4c)
  • Sync official Remotion skills (upstream v4.0.542, 1af7e69) (4b80ea2)
  • FEAT: templates/ds-release-notes takes the same DS act palette: flat DS blue title and outro with white type, cool grey item cards and stage with navy type, no gradient or shadow on the cards (45db3d0)
  • FEAT: examples/ds-release-649 takes the DS act palette from video-design: title and outro on flat DS blue with white type, item cards on cool grey with navy type, no gradient, and larger type; the video on demos is re-rendered (d79c6f7)

Full Changelog: v0.27.0...v0.28.0