You already know the account. This is the sharpening — Aldo's own revision notes turned into rules you can apply to any clip, so the next cut gets approved on sight.
"Ask yourself: do you think this looks real? If you wouldn't show it to your friends and tell them it's real — it's not done."
Each of these is a repeated note from Aldo, generalized into a principle. Don't apply them literally to one clip — apply the thinking to every clip.
A clean render reads as CGI. Real phone footage has grain, motion blur, compression, murky low light. Blur and rough-up every AI/3D shot until it could pass for something a camera actually caught.
"Blur the whole thing — everything has to look real."
Someone is "holding" every shot, and they're inside the scene. Panic, running, a jump-scare → the frame shakes, whips, destabilizes. A locked-off shot during chaos looks like CGI on a tripod. Match the camera's motion to the adrenaline.
"Add a handheld shake — we're freaking out, the camera wouldn't be still."
A muted AI clip feels pasted in; its native audio grounds it in reality. Keep the generated/source audio on, then mix it under the track — don't discard it.
"Don't mute the audio — there's a setting on Higgsfield to keep it on."
The join between real and AI is where the eye catches the fake. Mask every transition by cutting to a real POV shot first, then reveal the AI — and trim out any generation glitch entirely.
"You can see the transition into AI — switch to US to mask it."
Drive the AI from his actual footage, not a blank prompt. Match the real vehicle, location, physics — and whatever he's narrating. If he says everything's getting destroyed, the frame shows destruction.
"I'm in a Ferrari not a Tesla — use the footage you have."
Never let an AI-generated face that isn't Aldo read as a person on screen. Cut before it lands — it's him, or it's gone.
"That's obviously not me — delete as soon as you see my face."
Every second of story earns a visual. Reaction, article-reading, and storytelling stretches get imagery, popups, AI or effects layered over them. A bare talking head is "boring."
"No imagery while they're telling the story — the first 10 mins is boring."
Trim the silence and breathing gaps — keep momentum. On drone beats, don't sit on one angle: switch back and forth between the full view and the drone split-shot.
"Cut the dead spots so it feels more fast-paced."
Music isn't wallpaper. Time it to mood, and drop to suspense SFX with no music on the creepy beats. Knowing when to pull the music is as important as the track itself.
"You gotta know when to use music."
Voices, 3D models, animation layers are there to be used and combined — turn the "hair attack" and "idle" on together, drop in the screen-recorded voice. Inventory the folder; leave nothing on the table.
"There's a hair attack — why didn't you use it? Bring the character to life."
When he sends an inspo link, study it — pull its SFX, copy its rhythm, compare your output against it. The reference is the target, not a suggestion.
"Study their use of music & practice on your project."
His feedback is always about one clip — but there's a general rule underneath every note. Train yourself to make this jump, and you'll fix the whole video instead of the one timestamp.
Add a handheld shake to this clip. We are freaking out right? The camera wouldn't be still, it would be shaky.
Make sure to not mute the audio! There's a setting on Higgsfield to keep the audio on.
If you wouldn't show your friends and say it's real, blur it.
I'm inside a Ferrari not a Tesla — use the footage you have, not this.
No imagery while they're telling the story — the first 10 mins is boring.
These are the notes that have come back before. Run the list on every cut before you post the Frame link — catching them yourself is the difference between approved-after-revisions and approved-on-sight.
Would your friends believe a real camera caught this — shaky hands, real sound, not one dead second? If not, you're not done yet.