Blocked Lists, Feed Control, and Cleaner AI Video Workflows

Most people open their blocked list exactly once, during a bad week, and then never again. That is a mistake, though not the one you expect. The list is not a record of grudges. It is a record of decisions about visibility, and visibility decisions quietly shape both what you see and what you publish. The same logic applies the moment you sit down to generate video: the clips you ship are defined as much by what you exclude as by what you ask for.

This guide handles both halves of that idea. First comes the practical navigation — where the blocked list actually lives, what changes when you block someone, how to reverse it, and how to choose between muting, restricting, blocking and reporting. Then comes the transfer: how the exclusion mindset, applied to negative prompts, reference triage and export checks, turns chaotic generation sessions into a repeatable workflow you can trust.

The blocked list is a control surface, not a grudge list

Blocking usually happens in an emotional moment, which is why people think of it as an emotional act. Functionally it is a settings change. You are telling a platform that one account no longer gets to see your profile, contact you, or appear in your interactions. That is a visibility decision, and visibility decisions compound the way interest does.

A creator account with forty blocks spread across three years is telling a story about its boundaries. An account with zero blocks after a harassment spike is telling a different story. Neither is automatically wrong, but both deserve a deliberate review rather than an accidental one.

Reframing the list this way makes it easier to notice the parallel in AI video work. Most people treat prompt writing as the creative act and everything else as housekeeping. In practice the housekeeping carries most of the quality. A short negative prompt that removes warped hands, flickering frames and stray text does more for a finished clip than another twenty adjectives piled into the positive prompt. A reference folder with permission notes prevents the licensing problem that surfaces two weeks after publishing, usually on the day a client asks where a background came from.

Before you touch any menu, decide which question you are actually asking:

• Auditing access. Who can reach me, tag me, or message me right now?
• Cleaning a feed. Why does my For You page keep surfacing a genre I scroll past?
• Verifying a change. Did the block I set last month actually take effect?
• Recovering a relationship. Someone I blocked is now relevant again and I want the door open.

Each question routes to a different screen. Mixing them up is the main reason people conclude the setting does not exist.

Finding the blocked list: the path that actually works

The blocked list lives inside privacy settings, which is why so many searches end in the wrong menu. Notification preferences, content preferences, screen-time tools and safety settings all sit nearby and all look plausible at a glance. Here is the reliable route.

On the mobile app

Open your profile, tap the menu icon in the top corner, and choose Settings and privacy. Open Privacy, then scroll to the section that governs how other people interact with you. The blocked entry sits in that group. On most versions the label reads "Blocked" or "Blocked accounts," and tapping it opens a flat list of avatars and usernames with an unblock action beside each row.

Three practical notes save real time here. First, the list has no built-in search, so if you are hunting one specific account you scroll rather than type; if the list is genuinely large, screenshot it in chunks so you can review it later without reopening the menu. Second, labels shift between app versions and languages, so an unfamiliar translation is not evidence that the feature vanished. Third, business and personal profiles do not always expose the same settings — if you manage both, check each separately.

On the desktop web version

The same list exists in the browser, but it is buried deeper, typically inside the account centre under privacy. People who switch devices mid-task often assume the list disappeared because they cannot find it in the mobile menu they are used to. If a colleague insists the option is gone, ask which surface they are on first.

When the menu does not match what you expect

The most common cause of a mismatched menu is a stale layout, not a removed feature. Settings get regrouped during updates, translated inconsistently, and sometimes rolled out to a fraction of accounts before everyone else. If your navigation looks nothing like the description above, search the in-app help centre for "blocked accounts" rather than tapping through every submenu. The help centre reflects the current layout of the build you are running, while a tutorial screenshot from last year does not.

What blocking changes, and what it does not

Blocking is stronger than most people assume and weaker than some fear. Once you block an account, that person generally cannot view your profile, follow you, comment on your videos, or send you direct messages. Existing comments from that account typically stop appearing in your view. On their side the relationship is severed in both directions, so they usually cannot find you through search either.

The side effects matter more than the headline. Because blocking is reciprocal in visibility, it can break shared contexts: a group chat link, a collaboration thread, a mutual friend's comment section, a live stream you both happen to join. Blocking also does not erase anything already screenshotted, downloaded or reposted elsewhere, and it does not stop a determined person from opening another account. That is why the list is a hygiene tool rather than a security system.

Unblocking is similarly partial. Tap the unblock action and the restriction lifts, but the relationship does not reset to what it was. You will not automatically follow each other again, old conversations do not reappear, and nobody is notified that a block existed at all. If you want contact restored, you follow them again and wait.

A sensible rule: unblock only when you have a reason that would survive a bad day. Releasing a block after an argument cools is fine. Releasing it because the list makes you feel guilty is how the same problem returns a month later.

Mute, restrict, block, or report: choosing the lightest tool

Blocking is the loudest option available, and it is rarely the correct first move. Most platforms give you a graduated set of controls, and matching the tool to the actual problem prevents a lot of collateral damage.

Situation	Best control	Why it fits
Someone's content grates on you but you want no conflict	Mute	No notification, no severed visibility
One account keeps leaving disruptive comments	Restrict	Comments held for review without an announcement
You want zero contact and zero visibility	Block	Cuts both directions
A community guideline is genuinely broken	Report	The only path that triggers review
Your feed has drifted toward content you dislike	Behaviour plus "not interested"	Ranking responds to behaviour, not toggles
AI output keeps producing the same defect	Negative prompt defaults	Fixes the cause rather than one clip
Reference images of unclear origin	Asset folder triage	Prevents licensing exposure later
Exports keep failing a platform check	Output checklist or template	Removes per-project guesswork

The pattern is consistent: pick the lightest control that solves the real problem, then make that control a default so you never solve the same problem twice.

Why the recommendation system ignores your settings

A blocked list and a feed are not the same system, and treating them as one causes endless frustration. Blocking removes a specific account. Recommendation ranks content by signals you never explicitly set — watch time, replays, shares, saves, and the speed at which you scroll past.

The asymmetry explains a very common experience. If you watch a genre to the end out of disbelief, the system reads engagement. Hatred with attention still looks like attention. A single fast scroll is weak evidence against a pattern you keep rewatching.

Training the feed with behaviour

The fix is behavioural rather than declarative. Use "not interested" on the specific videos that represent the pattern. Stop rewatching material you want less of, even ironically. Give the system clean positive signals by finishing the content you actually want more of. Over a week or two, that combination outperforms any settings menu, because it is the signal the ranking model was built to read.

Keyword filters and restricted modes

Most platforms now offer keyword filters for comments and, in some cases, for recommended content. These are badly underused. A short list of slurs, spam phrases and recurring scam patterns catches a surprising share of noise before it reaches you, and it costs five minutes to set up once.

Restricted modes are the other lever. They limit how much unknown content surfaces, which is worth enabling when a feed has drifted somewhere unhelpful or when you hand a phone to someone else. Neither replaces blocking, but both reduce the number of situations where blocking becomes necessary.

From platform controls to production controls

Here is where the blocked-list mindset earns its keep. An AI video project has three layers where exclusion matters: the prompt, the reference assets, and the final output. Treat each one as a place to stop problems before they enter, instead of a place to repair them after they appear.

Exclusion at the prompt layer

Negative prompts are the blocked list of generation. Rather than describing what you want more loudly, you name what must not appear. Useful defaults across most models cluster into four groups:

• Anatomy and motion: extra fingers, warped hands, duplicated limbs, melting facial features, feet sliding across the floor.
• Rendering artifacts: burned-in text, watermarks, logos, jitter, frame flicker, over-sharpened edges, colour banding.
• Composition errors: cropped heads, off-centre subjects, floating objects, shadows pointing the wrong way.
• Style drift: cartoon shading inside a photoreal shot, plastic skin, oversaturated colour, inconsistent film grain.

Keep the list short and specific. Twenty generic words dilute the signal; five to eight targeted terms, adjusted per shot, work far better than a master list pasted into every field. If you want a starting structure, the prompt library is a reasonable place to borrow phrasing rather than staring at an empty box.

Exclusion at the asset layer

Reference images and style frames are where licensing and consent problems enter a project. Before a frame becomes a reference, answer three questions: do I have the right to use this, does it contain a recognisable person, and would I be comfortable explaining the source on camera?

A simple folder convention handles most of it. Keep three directories — cleared, pending, rejected — and move files between them deliberately. Nothing in the pending folder enters a generation session. This feels bureaucratic until the first time a client asks for provenance and you can answer in ten seconds. With the AI image generator, the same discipline applies to style frames: lock a look you have rights to, then animate from it.

Exclusion at the output layer

Publishing is the last place to catch mistakes and the cheapest. A short pre-export checklist does the job:

1. Watch the clip once at full size with the sound off, then once at thumbnail size.
2. Check the first and last two seconds — artifacts cluster at boundaries.
3. Confirm no unintended text, logo or brand mark survived in frame.
4. Verify aspect ratio and safe margins for every destination you plan to publish to.
5. Confirm music and voice rights, and disclose synthetic voice where your audience or client expects it.

If you produce several variants per week, templates keep that checklist consistent instead of re-deriving it each time. The video templates collection is built around exactly that idea: start from a structure that already respects the boring constraints.

A worked example: a thirty-second product teaser

Abstract advice is easy to nod at and hard to apply. Here is how the layers stack on a realistic brief: a thirty-second teaser for a ceramic coffee mug, built for vertical feeds, with no budget for a shoot.

Step one, define the exclusions before the scene. The brief says clean and premium, so the negative list starts with burned-in text, watermark, distorted hands holding the mug, steam that looks like smoke, plastic-looking glaze, and harsh specular highlights that read as cheap. Five to eight terms, chosen because they describe the failure modes of this specific product category. Ceramics fail in particular ways: glaze looks plasticky, rims warp, shadows turn muddy.

Step two, clear the references. Two photos from the client's own catalogue go into the cleared folder. One image found in a search results page goes into rejected, because nobody can confirm its origin. A third frame the client wants "something like" goes into pending until written permission arrives. No generation session touches pending files, which means no late-night accidental inclusion.

Step three, generate the safe shots first. A slow push-in on the mug on a wooden surface. A pour shot where the hand is out of frame. A rack focus from rim to handle. Each shot is short — three to five seconds — because short clips hide the continuity problems that long AI shots amplify. Each generation uses the same negative defaults, plus one shot-specific addition, such as "no dripping liquid on the table."

Step four, review at two sizes. Full screen with sound off catches composition and artifacts. Thumbnail size catches the thing that actually matters in a feed: whether the subject reads at a glance. Three of eight shots fail at thumbnail size because the mug occupies too little of the frame, and the fix is framing, not prompt length.

Step five, standardise what you learned. The three failures become defaults: subject occupies at least forty percent of frame height, no wide establishing shot for a single product, no camera move faster than a slow push. Those defaults go into the template for the next product, which cuts the next session roughly in half.

That is the whole method. Exclusions at the prompt, triage at the assets, checks at the output, and every lesson promoted into a default so it is never relearned.

Mistakes, audits, and defaults that hold

Control systems decay. A routine beats willpower, and defaults beat both. Twenty minutes once a month is enough if you run the same sequence in the same order.

Minutes zero to five: read the blocked list. Open it, read every name, and ask one question per row — does this still need to be here? Unblock the ones that no longer serve a purpose, and note any account you blocked in anger that you would rather have muted.

Minutes five to ten: check adjacent settings. Comment filters, restricted mode, message permissions, tag permissions. These drift after updates, and a filter you enabled last quarter may have been reset or regrouped under a new heading.

Minutes ten to fifteen: audit project hygiene. Open the reference folders, delete stale drafts, and update negative defaults based on what actually failed last month. If you keep a failure log — even a plain text file listing shots that needed three attempts — this is the moment it pays for itself.

Minutes fifteen to twenty: review output. Pick two recent clips and watch them the way a stranger would. Note one recurring flaw. Fix that flaw in your defaults rather than in the next individual prompt. Fixing a defect once in a template saves you from fixing it forty times.

The mistakes that quietly cost hours are consistent across both halves of this guide:

• Blocking when muting would do. Blocking severs visibility both ways and can break shared contexts. If the goal is a calmer feed, mute and move on.
• Assuming a block hides everything. It hides your profile and interaction, not screenshots or reposts. If real harm is involved, reporting and platform safety tools matter more than a block.
• Treating the algorithm as a settings menu. There is no toggle that says "stop showing me this genre," but there is a pattern of watch behaviour that does exactly that over a fortnight.
• Pasting an enormous negative prompt. Long exclusion lists create conflicting signals and can flatten an image into mush. Specific beats exhaustive.
• Skipping the rights check on references. The most expensive mistake in AI video is not a bad render. It is a beautiful render built on a frame you cannot legally use.
• Never reviewing the list. A blocked list you have not opened in a year is not a boundary. It is an archive of unsupervised decisions.

FAQ

Does the person know when I block or unblock them? No. Neither action generates a notification. They notice the effect — your profile stops resolving, comments stop appearing — but not the event itself.

Where do accounts go after I unblock them? They leave the list entirely. There is no separate archive of previously blocked accounts, so if you want a record for your own notes, keep it yourself.

Can other people see my blocked list? No. The list is private to your account regardless of their settings or yours.

Does blocking delete old comments? From your perspective, comments from that account generally stop appearing. That is not the same as deletion, and anything already captured elsewhere stays captured.

Can I block a topic instead of an account? Not directly. Keyword filters cover comments, and "not interested" plus watch behaviour shapes recommendations. In AI work the equivalent is a negative prompt library that encodes the motifs you never want to see again.

Why does my feed still drift if my blocked list is well maintained? Because ranking and blocking are separate systems. Blocks remove specific accounts; ranking responds probabilistically to engagement. A clean list and a calm feed are related, but they are not the same setting.

How many negative prompt terms is too many? If you cannot say why a term is there, it is too many. Start with five to eight that map to real failures you have seen, then add one shot-specific term at a time.

Should the same exclusions apply to every model? No. Different models fail differently, so keep a shared core of anatomy and artifact terms, then a per-model addendum based on observation rather than reputation.

Build a workflow where exclusion is the default

The blocked list is a small setting with a large lesson: control is mostly subtractive. Remove the accounts, keywords and visual artifacts you do not want, and what remains looks intentional. Creators who internalise that stop chasing the perfect prompt and start maintaining cleaner inputs — cleared references, tight negative defaults, a checklist they actually run.

That is the workflow the AI video generator is designed for: a cinematic idea, an explicit set of exclusions, and a render that respects both. Describe your scene, state what must not appear, and iterate on short clips rather than gambling on long ones. When you are comparing tools, the alternatives overview lays out approaches without the hype, and Orelon vs Runway digs into where each one fits a production pipeline. For deeper walkthroughs on prompt hygiene, reference management and export checks, the Orelon blog is the place to keep reading.

Clean lists, clean prompts, clean exports. That is the whole system, and it compounds every time you ship.