ext/hyperframes
2
0
Fork 0
mirror of https://github.com/heygen-com/hyperframes.git synced 2026-10-30 09:16:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(skills): signed in, Reverse remakes camera and generated pixels with HeyGen by default

Browse Source 
Presenters re-shot, stills and shots generated, music generated; cutting from the original is the offline
path, or what the person asks for.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Miao Yang
2026-10-28 22:56:26 +08:00
co-authored by Claude Opus 5.5
parent 604fb329e5
commit 2f11a5a72c
3 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
skills-manifest.json
+1 -1
View File
@@ -26,7 +26,7 @@
"files": 76
},
"reverse-any-video": {
"hash": "4501a1462aa399a2",
"hash": "98a05bbc5d584d10",
"files": 6
},
"slideshow": {
skills/reverse-any-video/SKILL.md
+4 -4
View File
@@ -36,14 +36,14 @@ Write `REVERSE.md` (`references/spec.md`): the brief, the recurring objects, the
## 2. Rebuild by layer
| Layer | Default | When the person asks for new pixels |
| Layer | Signed in (default) | Offline, or asked: local only |
| --- | --- | --- |
| Text, shapes, UI, charts, transitions | HTML + GSAP, registry first | — |
| Burned-in captions | a caption track styled like the original, timed from the transcript | — |
| Logos | SVG geometry; cut from the original when photographic | — |
| Photos, footage, AI shots | each cut on its own (`measure extract --roi`) | generated (`/media-use`) |
| A presenter | re-shot with HeyGen from a still and their own voice | — |
| Sound | the original's track (`measure extract --out assets/sound.m4a`) as one `<audio id>` | voice, music (`/media-use`) |
| Photos, footage, AI shots | generated from its frame (`/media-use`) | cut on its own (`measure extract --roi`) |
| A presenter | re-shot with HeyGen from a still and their voice | cut from the original |
| Sound | original voice (`measure extract --out assets/sound.m4a`), generated music | the original's track |
`references/layers.md` covers placement, burned-in text over reused footage, fonts and regeneration. One sub-composition per segment (one scene when a single camera spans several); each string one element bound to a variable; colours as `:root` variables.
skills/reverse-any-video/references/layers.md
+1 -1
View File
@@ -46,7 +46,7 @@ The original's sound is one file, `measure extract .reverse/source.mp4 --out ass
## New pixels
Presenters are re-shot by default when HeyGen is available; anything else is regenerated only when the person asks. Regenerated media is made with `npx hyperframes media-use generate` (`/media-use`). Ground every generation in the original:
Signed in, every camera or generated pixel is remade with HeyGen by default: presenters re-shot, stills and shots generated, music generated to the original's length and tempo. Cut from the original only offline or when the person asks for local only. Regenerated media is made with `npx hyperframes media-use generate` (`/media-use`). Ground every generation in the original:
- **A still:** `--type image --reference <the original frame>`, asked to keep the composition and leave out what HeyGen should not draw: a clean plate.
- **A shot:** `--type video --first-frame <the shot's first frame>` at the shot's length, 5–15 s; a longer shot is chained shots, each starting from the last frame of the one before (`measure extract --at` cuts both).