Video & audio

Video production pipeline

The whole chain: aspect lock, script, character consistency, keyframes, voice, assembly.

Paid $29 list, $23 until 2026-12-31 · one skill, yours for good. The outline, the steps and the file open after purchase.

Background

This is the orchestration layer: it only decides "what gets done when, how the parallel work is scheduled, and what condition counts as passing". The actual craft is delegated to existing skills and to this skill's scripts.

Why it is built this way

The first two questions are ordered by what changes what. The approved amount comes before the aspect ratio, because the amount decides the ratio, the duration and the shot count, and planning at a ratio the budget cannot pay for ends in rework. The ratio then gets locked before any image exists, written down with pixel values, because a generation call that omits the size takes the model default: one film came back mixing 3:2 and 16:9 frames in a 752×560 master that claimed to be portrait.

Stages serialise where one output feeds everything downstream, which is why the S0 four tables freeze: verbatim lines, the look-to-voice match, measured vocal onsets, the locked ratio. They parallelise by entity so consistency survives, anchors running across characters at once and within one character in order. Gates live in scripts because prose does not execute. First-and-last-frame chaining only took effect once it was written into the external brief, after eleven shots were generated independently, the costume changed four times and one bench became four places.

Where it bites

• A 3:2 first frame A generation call without --size hands the ratio to the model default, and 2496×1664 is the tell. check_aspect.py compares every image and clip against the locked ratio and stops the run on the first mismatch; catching 752×560 there costs one render, catching it after the keyframes costs the film.
• Eleven shots generated on their own Independent generation gives four costumes and four versions of the same bench while ratio, duration, lines and loudness all pass. Gate 3 requires first-to-last frame chaining written verbatim into the task brief and checked per shot, with two shots reviewed side by side before the rest.
• Trial runs on the main tier Testing composition on video instead of images is where the budget goes: one 5-second 1080P clip costs about 94 images, and the tier gap reaches 37 times. The cost gate runs --dry-run before every call, verifies composition on images, and puts technical tests on the 480P mini tier.
• Retries and steps billed apart enable_retry bills each retry separately, a two-step task bills each step, and image grids bill per image produced, so a batch can look cheap and land far over plan. Every charge is booked the moment it happens, a rejected call is booked at zero, and 10 percent over plan stops the run.

When to use it

End-to-end video production pipeline (orchestration layer): aspect-ratio lock → script and storyboard → character consistency → keyframes → image-to-video → performance and lived-in feel → voiceover, music and sound effects → code-driven compositing → QC and delivery, supporting landscape 16:9 and portrait 9:16, with digital-human likeness compliance and false-positive appeals plus on-screen common-sense verification. Use when the user asks to make a video / short drama / concept film / commercial, to turn a piece of text into a video, to get a landscape or portrait final cut, for a video production process, or to lock the aspect ratio/proportion.