Grain, flare and vignette against the AI sheen
Generated footage reads wrong because imperfection is absent, not detail. A finishing order for grain, vignette and flare, and the amounts that hold up.
9 min read
The hand deforms as soon as it holds the product
Contact with a product is where hands fail. A grip-region mask-and-inpaint recipe, then a composite fallback that keeps the real SKU geometry intact.
9 min read
Hard or soft: source size in lighting prompts
Shadow edge hardness is set by apparent source size, not by adjectives. A size-and-direction vocabulary that outperforms five style words in every prompt.
9 min read
How Meta decides to show an AI Info label
Meta's AI Info label fires from industry-standard file signals or from self-disclosure. Which of those you control, and why a real photo can still get labelled.
8 min read
How to read a GPAI training-data summary
GPAI providers must publish a training-content summary. Six questions that separate a useful one from a decorative one, and what the document cannot prove.
8 min read
Identity blending: a face that isn't anyone
Weighted mixes of faces a model already knows produce a novel character that reproduces. The syntax, the ratios that stay stable, and how to lock the result.
9 min read
The sign is misspelled. Fix it without re-rolling
Do not re-roll a good image for one misspelled word. Quote the exact string, describe the font, set a mask margin, and know when to composite type instead.
8 min read
Instagram Reels cover photo spec and crop
Instagram caps Reels covers at 8MB and centre-crops anything not supplied at 9:16. Here is the crop arithmetic and a cover template that survives it.
8 min read
Instagram Reels encoding: codecs and frame rate
Reels take MOV or MP4, H.264 or HEVC, 23 to 60fps. Stay under three minutes if you want recommendations, including from a 25fps timeline.
8 min read
Interrupting an agent mid-turn
Answer now stops the reply from launching more tools. Cancel stops a background task. Closing the tab aborts the turn without unsending work already dispatched.
7 min read
Isolate one variable in short-form tests
If caption, sound, hashtags, and hook all differ, the result is unreadable. Use a variant spec sheet and lock the confounds people forget.
9 min read
J-cuts and L-cuts with generated audio
Stitched clips read as a slideshow because picture and sound change on the same frame. How to split those edges when the audio came out of the generation.
9 min read
Karaoke captions highlight the wrong word
Word-by-word caption styles expose interpolated timestamps. Force-align with a phoneme model before you style, or the highlight sits on the wrong word.
9 min read
Karaoke captions vs static blocks, by format
Word-by-word burn-in holds in short-form and fatigues in long-form. The per-format rule, and how to catch caption-style cost at specific timestamps.
9 min read
Keep the prompt structure, vary only the action
Multi-shot consistency comes from freezing the prompt skeleton and changing one clause. The template, the fields that must never move, and how people break it.
9 min read
Keeping captions in sync with a dub
Captions burned before a dub drift against the translated audio. The fix is an order of operations that transcribes each dubbed track on its own.
9 min read
Keeping client material out of training data
Enterprise clients now require brand assets never enter a training set, and the duty flows through you to every tool. The clause, plus the audit behind it.
9 min read
Keeping reverb out of generated dialogue
Reverb is the most common uninvited guest in native audio. The room words that cause it, the exclusion phrasing that works, and what to do with a good take.
9 min read
Layout-first prompting for posters and covers
Some models now plan composition before they render. Prompt patterns that pin zones for posters, covers and thumbnails without dropping into coordinates.
9 min read
Writing license terms for prompt packs and LoRAs
Four decisions decide whether a buyer comes back: commercial use, redistribution, retraining, and derivative outputs. Written out as plain-language tiers.
9 min read
What LinkedIn's Content Credentials icon shows
LinkedIn's C2PA icon reveals whether AI was used, which tool made the file, and when the credential was issued. Here is what a click shows and what strips it.
9 min read
LinkedIn video upload limits and aspect ratios
Organic LinkedIn video is 15 minutes, 1:2.4–2.4:1, 256x144–4096x2304. Ads and Pages publish different sheets. One conservative export clears all three.
8 min read
Lip sync breaks on fast speech and plosives
Rapid consonants expose a 1–3 frame lag slower speech hides. Target a 120ms window, nudge the clip, and rewrite lines that keep tripping it.
9 min read
How to get a genuinely locked-off shot
Leaving camera language out of a video prompt does not give you a static frame. The explicit phrasing and staging that produce a truly motionless camera.
8 min read