Local gym UGC is members who opted in
Local gym UGC is members who opted in. Film the floor, classes, and staff; do not mint a crowd or a body that did not join.
8 min read
Hotel UGC is the room you can book
Hotel UGC is the room you can book. Film those rooms; generate only captioned objection-handling, never a suite you do not sell.
9 min read
Pharmacy UGC is the box, not a patient
Pharmacy UGC is the box, not a patient. No medical talking heads. Pack plus a pharmacist you actually recorded if licensed.
8 min read
PLG SaaS UGC is a workflow, not a mascot
PLG SaaS UGC is a workflow, not a mascot. UI stills plus peer voice. Do not invent a CEO or a fake dashboard.
8 min read
Veo R2V is 8 seconds at 128 credits
VEO 3.1 Reference to Video is 8 seconds only, 128 credits silent or 256 with native audio, up to 3 stills. The stack is not a surcharge.
8 min read
Alt text is not a caption track
Alt text describes a still or post. Speech captions transcribe audio. Do not paste one into the other.
7 min read
Captions on carousel covers are a still
The first carousel frame is a poster. Burn a line into that still. Do not ship a timed track nobody plays.
7 min read
Stories captions sit higher than Reels
Story chrome (reply bar, stickers) eats the bottom. Place type higher than you would on Reels.
7 min read
Chapters are not a caption track
YouTube chapters are description timestamps. Burned-in lines are pixels. They are different objects.
7 min read
DYNAMIC captions on paid are a test
BASIC is the paid default. DYNAMIC costs 2× and is a variant, not the master. Keep a BASIC control.
7 min read
Forced narrative on pack shots
Translate the claim and the legal. Leave the trademarked SKU name. That leftover is forced narrative.
7 min read
Put the hook on screen before the SKU
Spend the first second on the spoken line, not the logo. Burn the claim, then show the pack.
7 min read
PIP the talking head over the product
Film or lipsync the face, I2V the product, then composite. One prompt cannot invent both plates.
7 min read
Dub a UGC ad without a new mouth
Lay a voice-over bed when the face is not the product. Lipsync only when the mouth has to match.
7 min read
Thumbnail from the keeper, not a new still
Extract a frame from the cut you will post. Do not text-to-image a face that is not in the video.
7 min read
Isolate vocals before karaoke captions
Strip the bed first. A dirty mix makes word-highlight land on the wrong syllable.
7 min read
Merge clips into one UGC file
Generate hook, demo, and CTA as three keepers, then merge. Do not prompt one take for all three.
7 min read
Replace the take's audio after TTS
Generate silent, speak with Gemini 3.1 Flash TTS (4cr), then replace the plate's audio. Do not mix.
7 min read
Trim dead air without a bad jump
Cut the breath on locked A-roll. Cover with B-roll if the face would pop on a blink.
7 min read
How to upscale a video to 4K
Lock the keeper, then run ByteDance (from 3 credits, up to 4K), Topaz (from 4) or Runway (from 8). Do not upscale a miss.
8 min read
Reading speed on a six-second ad
Count characters against six seconds. Cut words. Do not shrink type to fit a paragraph on a 6s spot.
8 min read
Speaker IDs on two-person UGC
When both people talk, label who is speaking. Auto captions will not invent SDH speaker IDs.
7 min read
A transcript is not an SRT
Plain text has no timestamps. An SRT is cues. Burn-in is pixels. Transcribe, then caption — different jobs.
7 min read
C2PA often dies on scheduler uploads
ImgKey, mid-2026: X strips C2PA; Graph API and cross-post paths are inconsistent; LinkedIn often preserves. Do not assume the manifest survived.
7 min read