Skip to main content

Auto vs Manual mode

When to let the pipeline run end-to-end and when to step in at the brief, the outline, the footage, or the publish surface.

5 min read

CueTheScene's whole pitch is that you can type a prompt, walk away, and come back to a publish-ready video. That's auto mode. You can also step in at any of four natural points and shape the video by hand, that's manual mode. There's no global toggle, the modes are per-job and per-stage. This page covers when each is the right call.

Auto mode

Default. Submit a brief, the pipeline does the rest, the next thing that needs your attention is the publish surface (and even that is optional, you can let it auto-publish to YouTube if you've configured it).

What "auto" means in practice:

  • The outline lands and proceeds to script without you reading it.
  • The script lands and proceeds to voice synthesis without you reading it.
  • The voice lands and proceeds to footage search without you previewing it.
  • The footage lands and proceeds to render without you swapping anything.
  • The render lands and the publish surface holds, waiting on your "Publish to YouTube" click. If auto-publish is on, that clicks itself.

Auto mode is the right call most of the time. The pipeline's quality gates (script critic, footage availability, VO anomaly, content policy) catch most things that need a human; they pause the job and surface a chooser panel rather than shipping a bad video. See "Paused for review" below.

Manual mode

Step in at any of four points. Each is the cheapest place to redirect the video without throwing away earlier work:

1. The brief

The biggest lever. Before you submit, the brief form has a "Help me write this" coach (rotating exemplars, niche-aware suggestions, a "polish what I've written" button) and a live "What you'll get" preview card on the right rail. If you tune the brief before submitting, you save credits and time vs regenerating later.

2. Post-outline

The outline is the cheapest stage to regenerate. Each section has a regenerate button with an optional "Sharpen with Claude" hint input where you type "make the opening more visual, cut the academic preamble" and Claude rewrites just that section. There's a sticky toggle for "Dry run" mode where you can iterate on the outline without spending credits on script generation; you commit to the script only when the outline reads right.

3. Post-footage

The footage approval page surfaces every scene's chosen clip, the candidates that lost, the vision model's confidence score, and a "Search more" affordance with source-tier filtering (All / Archival / Free stock / Pixabay). Per-scene one-click swap, per-scene regenerate. Confidence badges flag the scenes the vision model wasn't sure about so you can focus there.

4. Pre-publish

The publish surface has the title, the description, the tags, the thumbnail, the schedule, and the "Publish to YouTube" button. The AI's first picks are usually good. The editor lets you make them yours. Title-CTR predictor band, YouTube-style description preview, best-time-to-publish niche preset, all live.

Paused for review

A middle path between auto and manual. The pipeline runs auto until it hits a quality gate it can't resolve on its own, then it pauses and asks you what to do.

The four gates today:

Brief validator. Refuses generation that's likely to produce a bad video (hyper-specific visuals stock can't deliver, live-news asks, prompt-injection patterns). The chooser is "refine and resubmit" or "submit anyway".

Footage availability. Fewer than 70 percent of scenes have a vision-accepted candidate. The chooser is three cards: refine the brief and re-run, swap footage on the weak scenes by hand, or accept the downgrade and proceed with keyword-ranked fallbacks. Refine credits are free for the re-run.

VO anomaly. A scene's synthesised voiceover came back with an audible artefact (cutting out, slurring). The chooser is "regenerate just that scene" or "accept and continue".

Content policy. The script or footage flagged against the Acceptable Use Policy. The chooser depends on the category, usually it's "refine the brief" or "edit the script by hand".

A "Paused for review" pill on the dashboard means one of these is waiting on you. Click through, the chooser is the first thing you see.

When to choose which

Stick with auto when you're publishing on cadence (three to five videos a week), when the niche is one the pipeline knows well (history, science, finance, true crime, gaming), when the brief is well-formed, when you trust the pipeline to be steady-state.

Step in at the brief when the topic is unusual, when angle matters more than topic, when you have a strong opening or constraint you want preserved verbatim.

Step in at the outline when the niche is new to you or new to the pipeline, when you want to verify the angle before paying for the script.

Step in at the footage when the video is high-stakes (founding-member showcase, sponsored content, a video you're going to promote heavily), when the niche is archival-heavy (history, true crime) and you want to vet the source material.

Step in at publish every time. The title, description, and thumbnail are the levers that determine whether the video lands; the auto-picks are good, the edit-and-tune cycle takes two minutes and is worth doing.

What walking away looks like

You don't have to keep the tab open. Email notifications fire on each stage transition (configurable in Settings → Notifications). The render-complete email has the editor link in it; the publish surface is one click from there.

If you'd rather watch in real time, the dashboard live-updates without a refresh and the job-status page shows stage-by-stage progress with the rotating flavour copy.

The "Start another" loop

After your first auto-mode video clears stage three, the dashboard surfaces a "While you wait, start another" CTA card. The pipeline can run multiple jobs in parallel up to your plan's concurrency limit (1 on Starter, 2 on Creator, 3 on Pro, 5 on Scale), so the right move for cadence is usually to queue the next video while the current one's footage stage runs.

What's next

Editor overview covers the post-generation editor in detail, including the per-stage regenerate panels, the footage swap dialog, the caption-style and audio-mixer panels, and the publish surface.

Your first video walks through every choice the brief form offers, with prompting advice for each.

Cheers,
Carl