從真實影片案例學習 MiniMax H3 提示詞

Glenn Williams · @GlennHasABeard

A fixed-camera H3 experiment revises white ink in dark water so a sinking drop stretches into a ghost before spreading.

「take7｜AI動画ラボ」 · @KRhxLIrwOLxIjfU

A fictional cola ad comparison reports H3 as the only tested model to spell SORA COLA correctly on one line.

Fairu 🕵️‍♂️ · @Fairu_90

An Ocean Network template test uses H3 text-to-video and synced audio for an unusual ASMR concept with a difficult viewpoint.

きゅうす のAIとつれづれ · @Ai_Quse

A creator reports roughly 60W idle draw after an H3 run and closing ComfyUI, compared with about 40W after power-on.

たくぞー｜AI動画ツール実測レビュー · @tkz_aiart

A side-by-side workflow uses H3 gray-character previs to guide blocking and timing in a Seedance 2.5 live-action finish.

sorastg524 · @sorastg524

An H3 Max Turbo fantasy clip pairs a moonlit fox-spirit character with deep red lantern light.

ダルトワ★TV · @MireilleDartois

A video-reference test recommends lower-resolution dance guides for speed, while retaining detail for facial expressions and fingers.

ata · @atasrb

A same-seed Tokyo test reports 18.6 versus 8.1 minutes on DGX Spark, with reduced small-text and facial detail in QAD-v2.

taiti · @taitimusic

A vertical music promo uses H3 animation with Opus 5.5 and JIZURA text motion for a mobile-game presentation.

まぷて · @mapte_fireball

A creator builds a small app to trim long reference videos to selected sections under 15 seconds and reduce their resolution.

みと Visuals×AI · @mito_ai_la

An H3 city-sky action study tests gliding movement and camera paths among urban buildings.

Jiao Dong · @jd_markovchain

A Nuva Lab creator presents a shot-by-shot short made with FastH3 Omni and a creative agent over two afternoons.

Md Ismail Šojal 🕷️ · @0x0SojalSec

A local H3 clip demonstrates an Anya Taylor-Joy likeness using a published 155MB LoRA and no reference images.

tazmannner · @tazmannner

A creator’s H3 hires-fix node renders near 1MP before upscaling to 1080p, reporting 492 to 237 seconds versus native 2MP.

StarSource · @StarSourceH

A creator adds a self-speaking video reference to H3 and reports less exaggerated mouth motion at a higher 1.2MP setting.

Alessandro Fazenda · @adrfazenda

A local H3 experiment uses a Claude-created workflow for an extended staged conversation clip.

Guian Fang · @Enderfga

SolarRenderer uses H3-based control and distilled LoRAs to restyle engine structure passes with a reference first frame.

Kōrerotronix · @Amurinoble

An eight-second H3 Max title-card example uses a Transformers space setting for an animated franchise-style reveal.

zev · @zevoo

A creator combines a Suno song with H3 visuals in an extended music video whose preview features a dewdrop on grass.

PhotoX86 · @PhotoX86

A roughly 30-second H3 segmented-video attempt builds a comedy premise around finding a restroom.

robert. · @robrtcode

A surfer demonstration uses Pruna’s H3-based reference model to keep the subject and board consistent across changing angles.

Glenn Williams · @GlennHasABeard

A fixed low camera frames a corn row while a short prompt isolates invisible bootprints; the creator describes correcting their direction.

Alice The Ai Expert · @AliceInfoAi

A labeled three-model collage comparison contrasts oversized letters, cutouts and eyes, with MiniMax H3 in the middle panel.

Higgsfield AI 🧩 · @higgsfield_ai

A labeled split-screen compares rotating feathered wings in H3 and Claude Opus 5.5, focusing on detail, lighting and depth.

びおめた · @malariadx

A ComfyUI example uses a TextGenerate node with an H3-oriented system prompt and pairs its workflow screenshot with an anime clip.

Glenn Williams · @GlennHasABeard

A locked window shot defines a handprint as untouched glass while frost spreads from one corner in one direction.

buchi_ · @mix_buchi_

An anime-action test documents the creator’s concerns about limited motion and stereo width in an eight-step FastH3 Omni result.

まぷて · @mapte_fireball

A shared-reference, prompt and seed comparison reports 301s for Omni and 313s for LightX2V Ref2V Turbo at 1024×576.

🟢DeepForest_Studio_AI🟢 · @DeepForest_Inc

A five-second R2V test compares base H3, Omni, Kitchen and VSA, reporting faster renders alongside facial artifacts.

EasyChen · @easychen

A creator’s 22-shot short-drama study reports improved voice similarity and faster-motion clarity, with wardrobe leakage onto extras.