Anil-matcha/vox-ai-motion-graphics-generator
muapi-director
VerifiedTurn ONE topic, talking-head video, or photo into a finished Vox-style paper-collage explainer / ad video on the MuAPI platform (api.muapi.ai) + local ffmpeg — script, collage keyframes, motion, voice-over, music, captions, all automated. Three input modalities: a topic (B-roll), a talking-head video (A-roll mode), or a single photo of a person/product anchored into the collage (C-roll mode).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.7 KB |
| assets/demo.mp4 | binary | 6.3 MB |
| assets/youtube-thumbnail.jpg | binary | 100.1 KB |
| examples/sample-beats.json | config | 2.4 KB |
| references/beat-layer.md | markdown | 3.3 KB |
| references/prompt-guide.md | markdown | 3.0 KB |
| references/voices.md | markdown | 1.4 KB |
| scripts/aroll_assemble.py | script | 2.6 KB |
| scripts/aroll_clips.py | script | 4.5 KB |
| scripts/asr_beats.py | script | 4.9 KB |
| scripts/assemble.py | script | 6.4 KB |
| scripts/audio.py | script | 3.7 KB |
| scripts/clips.py | script | 5.8 KB |
| scripts/croll_keyframes.py | script | 5.1 KB |
| scripts/keyframes.py | script | 3.1 KB |
| scripts/muapi_client.py | script | 9.0 KB |
| scripts/provider.py | script | 5.5 KB |
| scripts/style_bakeoff.py | script | 2.4 KB |
| scripts/styles.py | script | 11.5 KB |
| scripts/text_overlay.py | script | 4.6 KB |
scripts/croll_keyframes.py
View on GitHub#!/usr/bin/env python3
"""
C-roll keyframe stage: anchors ONE still photo inside generated collage posters ("cutout roll").
"""
import json
import os
import sys
from provider import get_provider, run_jobs
from styles import compose_collage_prompt, resolve_theme, image_params
EDIT_MODEL = "flux-dev"
FACE_LOCK = (
"The person's face and hair from the attached photo are cut out as a PHOTOGRAPHIC "
"sticker with a torn white paper border — keep the facial identity, features and the "
"exact expression from the photo pixel-faithful; do not redraw, repaint or stylize the "
"face; NO halftone dots, print texture or ink treatment on the face or hair — the face "
"stays a clean photographic print. All poses and gestures are expressed by the body "
"only. From the neck down the body is a hand-drawn paper-doll illustration jointed "
"like a vintage paper puppet with visible cut edges, FULLY CLOTHED in {wardrobe}. "
)
PRODUCT_LOCK = (
"{subject} from the attached photo is cut out as a PHOTOGRAPHIC sticker with a clean "
"scissor-cut edge and a soft real paper drop shadow — keep its exact shape, materials, "
"surface reflections and every word of its label typography pixel-faithful. Do not "
"redraw, restyle or repaint the subject or its label. Re-style ONLY the world around "
"it as printed paper collage. "
)
CROLL_GUARDS = (
" Halftone dots and print textures live on the BACKGROUND only. Newspaper scraps carry "
"completely UNREADABLE blurred micro-text. No readable text anywhere in the image."
)
FREEZE = {
"portrait": ("FREEZE the photographic face sticker — it is a frozen layer, "
"pixel-identical to the still for the entire duration; never redraw, warp "
"or animate the face; the paper-doll body may shift slightly at its joints."),
"product": ("FREEZE the photographic subject sticker and its label — a frozen layer, "
"pixel-identical to the still for the entire duration; every letter of the "
"label stays exactly as in the still; it may only settle gently with its "
"drop shadow."),
}
def build_lock(doc):
kind = doc.get("croll_subject", "portrait")
if kind == "product":
return PRODUCT_LOCK.format(subject=doc.get("subject_desc", "The product")), kind
wardrobe = doc.get("subject_wardrobe", "the same outfit as in the photo")
return FACE_LOCK.format(wardrobe=wardrobe), kind
def shots_of(beat):
if beat.get("shots"):
for s in beat["shots"]:
yield s, f"{beat['id']}{s.get('id','')}"
else:
yield beat, f"{beat['id']}"
def run(project_dir):
bpath = os.path.join(project_dir, "beats.json")
with open(bpath) as f:
doc = json.load(f)
photo = doc["anchor_photo"]
aspect = doc.get("aspect", "9:16")
img_model = doc.get("image_model", EDIT_MODEL)
img_res = doc.get("image_resolution", "1k")
theme = resolve_theme(doc.get("theme")) or {}
collage_style = theme.get("idiom") or doc.get("collage_style", "american-retro")
t_palette = theme.get("palette") or doc.get("palette")
t_type = theme.get("type_style") or doc.get("type_style")
t_finish = theme.get("finish") or doc.get("finish")
lock_text, kind = build_lock(doc)
doc["anchor_freeze"] = FREEZE[kind]
prov = get_provider(doc.get("provider"))
if not photo.startswith("http://") and not photo.startswith("https://"):
photo_url = prov.upload(photo)
doc["anchor_photo_url"] = photo_url
else:
photo_url = photo
kf_dir = os.path.join(project_dir, "keyframes")
os.makedirs(kf_dir, exist_ok=True)
specs, by_key = {}, {}
for beat in doc["beats"]:
for shot, key in shots_of(beat):
if shot.get("keyframe_url"):
continue
scene = shot["scene"]
base_prompt = compose_collage_prompt(scene, "", "", beat.get("bg", "warm ochre"),
aspect, with_title=False, style=collage_style,
palette=t_palette, type_style=t_type, finish=t_finish)
prompt = f"{lock_text} {base_prompt}{CROLL_GUARDS}"
shot["keyframe_prompt"] = prompt
params = {"image_url": photo_url, **image_params(img_model, aspect, img_res)}
specs[key] = (lambda p=prompt, par=params: prov.submit_image(img_model, p, **par))
by_key[key] = shot
done = run_jobs(prov, specs, poll_s=3, stall_s=75, max_retries=2, deadline_s=300)
for key, url in done.items():
if not url:
continue
dest = os.path.join(kf_dir, f"kf_{key}.jpg")
prov.download(url, dest)
shot = by_key[key]
shot["keyframe_url"] = url
shot["keyframe_path"] = dest
print(f"[{key}] saved {dest}")
with open(bpath, "w") as f:
json.dump(doc, f, ensure_ascii=False, indent=2)
print("Updated", bpath)
if __name__ == "__main__":
proj = sys.argv[1] if len(sys.argv) > 1 else os.path.join(
os.path.dirname(__file__), "..", "out", "croll-demo")
run(os.path.abspath(proj))
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |