@di-atomic/video-director

@di-atomic/video-director

Name
@di-atomic/video-director
Tagline
A shot list the renderer accepts unmodified, with arithmetic that does not ship black frames and voiceover lines that fit their shots.
Hero subhead
Transitions overlap their neighbours, so a plan summing to 60 seconds renders 57 seconds of content and 3 seconds of black. This solves for content instead of for the raw sum, quantizes generated clips to the lengths models actually produce, and times every voiceover line against the shot it plays over, from speech rates measured in 7 languages. It can also fill a template that video-analyst measured from a video you liked, and hand back what the engines cannot shoot.
Who it's for
Anyone whose agent makes video, and who has watched a render succeed and still come back wrong.
Version
0.3.0
Tier
beta
Path type
guidance-only
Install command
opvs-skills install @di-atomic/video-director
Trigger phrase
does this script fit 60 seconds
Marketplace URL
https://market.opvs.ai/v/di-atomic/video-director
Blog URL
https://di-atomic.com/blog/shipping-di-atomic-video-director
Stat 1 value
3.0s
Stat 1 label
of black a naive 60-second plan ships, measured on a real render
Stat 2 value
11 of 35
Stat 2 label
voiceover beats in my own 60-second demo scripts that ran past their shot
Stat 3 value
12 of 12
Stat 3 label
adversarial shot lists rejected, including an empty object
Stat 4 value
5
Stat 4 label
verifiers that print PASS or FAIL with numbers before a clip is bought
What headline
What you actually get
Cap 1 icon
🧮
Cap 1 title
Arithmetic that accounts for overlap
Cap 1 desc
Transitions overlap the shots either side of them, so content = sum(len) minus (n-1) times transition_seconds. Seven clips totalling 60 seconds with six half-second cuts render 57 seconds and 3 seconds of black. The planner solves for content, and the verifier prints the black tail you would have shipped.
Cap 2 icon
🎛️
Cap 2 title
Durations quantized to what models make
Cap 2 desc
veo3 returns 8 seconds, sora2 returns 10 or 15, and no model honours a request for 6.3. Screen recordings and motion compositions are elastic and absorb the remainder, so hitting a target is a small search rather than a subtraction.
Cap 3 icon
⏱️
Cap 3 title
Every voiceover line timed
Cap 3 desc
Every voiceover line is timed against its own shot, not the whole video, from speech rates measured for EN, DE, ES, FR, HE, RO and RU. A beat that does not fit goes back to the writer with the word count that does. --tts swaps the estimate for real timings.
Showcase headline
Engine is how the asset is made; type is derived from it
Showcase before caption
ai and avatar are generated, capped at 5 to 15 seconds and only on native lengths.
Showcase after caption
screen is a real recording and motion is a composition. Both are any length, and both are where the slack goes.
Worked title
A 60-second script that fits its total and fails 3 beats
Worked example
verify-vo-fit.mjs demo-script.md lang en, rate 2.38 (measured, conservative) 0-5s Hook window 4.9s est 4.62s PASS 5-15s Setup window 9.9s est 11.76s FAIL +1.86s max_words 23 50-58s compliance window 7.9s est 10.08s FAIL +2.18s max_words 18 58-60s Close window 1.75s est 5.46s FAIL +3.71s max_words 4 measured total speech 55.1s of 60s (the total fits) RESULT ........ FAIL (3 beats) --tts with the real voice: 2 real overflows, the setup fits verify-shot-list ......... PASS content 60.00s verify-seam-ledger ....... PASS 0 defaulted transitions
Worked framing
The total fit with five seconds to spare, and three beats still did not. Only those beats go back to the writer, each with the word count that fits, before anyone records a word.
Rel 1 icon
🎨
Rel 1 name
@di-atomic/media-generator
Rel 1 desc
Renders the ai and avatar rows this skill plans.
Rel 1 url
https://market.opvs.ai/v/di-atomic/media-generator
Rel 2 icon
🔍
Rel 2 name
@di-atomic/video-analyst
Rel 2 desc
Measures a video you liked into the template this skill fills.
Rel 2 url
https://market.opvs.ai/v/di-atomic/video-analyst
Rel 3 icon
🖊️
Rel 3 name
@di-atomic/copy-engine
Rel 3 desc
Writes the script this skill turns into shots.
Rel 3 url
https://market.opvs.ai/v/di-atomic/copy-engine
CTA headline
I got tired of renders that succeeded and were still wrong.
CTA description
video-director is one skill in the system behind Di-Atomic, the marketing and compliance agency that runs cognitoAI, SpiderIQ and OPVS. Guidance-only, so there is no backend to provision and nothing to wire up. It decides, composes and checks the clock; it renders nothing and writes no words, and it says so.
Upstream license
MIT