Direct AI fight scenes through controlled iteration: Replace game-like coverage with cinematic direction

2026-08-13 · Academy lesson · Adil · recording, 2:18 · original

Topic: Prompting & AI assistants (also: Video generation & shot direction) · mode: Tutorial · tags: course chapter, series, one creator's experience, prompts shared, fiction

In the last lesson of Direct AI fight scenes through controlled iteration, Adil shows why the full fight came out like a video-game cutscene (the negation trap and a camera floating behind the hero) and fixes it with direct descriptions instead of bans and four hard cuts with their own lenses, written by Claude.

Part of another source

This recording is almost entirely part of Hell Grind: Rebuilding Our Biggest Fight Scene from Scratch (2026-08-05); what is said in it is covered on that page.

What is said

• 00:00:00 — The full prompt and the first take of the fight
• 00:00:27 — The video-game look, the negation trap and the two fixes
• 00:01:19 — Four hard cuts with their own lenses, the final prompt from Claude and the result
• 00:01:55 — The lesson: tell the model what you want, run a few batches

Threads in this video

• restate · Should a prompt list what to avoid? — Adil's first fix for the video-game look: swap bans for direct descriptions, since the model gives what you ask for, not what you ban; describe the lighting and textures you want instead of "no cheap effects". (00:00:52)
• restate · How do you direct a multi-shot sequence? — Adil's second fix for the video-game look: make the model edit like a director with four hard cuts, each with its own camera setup: a 50mm orbit, a 24mm low angle, an 85mm close-up and a 35mm wide. (00:01:19)
• restate · How do you get a cinematic look in a generated image or shot? — Adil's second fix for the video-game look: make the model edit like a director with four hard cuts, each with its own camera setup: a 50mm orbit, a 24mm low angle, an 85mm close-up and a 35mm wide. (00:01:19)
• restate · Which LLM should be used to write video prompts? — For the four-cut version of the fight Adil sent one more request to Claude and got the final prompt for the definitive version of the shot. (00:01:19)
• restate · When a generation fails, should you generate more takes, fix the prompt or start over? — Adil's lesson for fight scenes: do not tell the model what not to do, tell it what you want, run a few batches and pick the best shots. (00:01:55)

Mentioned

Central: multi-shot prompt · negative prompt

Also: Claude · rebatching

Only in the description or credits: Direct AI fight scenes through controlled iteration · Higgsfield Academy

Also centred on negative prompt: Hell Grind: Rebuilding Our Biggest Fight Scene from Scratch

← Previous: Build a Brand's Visuals with AI: Packaging & Unboxing · Next → Make a Cinematic Ad End-to-End: Scenes 9-10: Round Two & The Winning Goal