Systematiq Ai

RUNWAY GEN-3 → APLHA VS TURBO. Quality of life tip… → If you want a quick clip use “Turbo”  → If you want more motion use “Alpha” Noticed people complaining about  Super “Slow-mo” results w. Runway Turbo. If you wait 45 extra seconds for your Gen… (With Alpha)...It pays off. MIDJOURNEY PROMPT:  bokeh motion, extreme closeup of a woman's face losing herself in the music, tulum jungle rave atmosphere, subtle neon light painting her face, 16mm, asymmetrical composition, moonlight, silhouettes, techno, --ar 5:3 --p 2wwavyd --stylize 250 --v 6.1 RUNWAY PROMPT: hand held shot of a woman cheering in a tulum jungle rave, silhouette lighting, 35mm style QUICK THOUGHTS:  + Turbo is a “speed” model + I use it for testing images to see if they work + Then I’ll use alpha for better results + You can’t use some of RW’s features with alpha + So if you want pure T2V outputs…use Alpha + You can also use first/last frame with Alpha + If you want to use features like… + Camera Motion/Keyframes…use Turbo #runwayml #midjourney #kling #minimax #ai #innovation 

WEEKEND PROJECT → AI DRAG RACE. With inspo from one Midjourney V4 image. Let the brain run wild with this spec ad. Using all the tools: + Midjourney + Magnific + Runway Gen-3 + Kling 1.6 + Minimax 01-Live + Sora + MMAudio >>This was purely for entertainment<< MIDJOURNEY BASE PROMPT: editorial photo, bmw on a road driving to the right, racing an F16 jet in the sky flying to the right, side view, speedvision, stromy skies, high contrast magazine photography, high precision photography with immaculate detail, sharp resolution --ar 2:1 --p --v 6 --q 2 VIDEO PROMOPT EXAMPLE: fast action trucking shot of a car racing an F16 fighter jet in the desert at 30x speed, they are both moving to the right at high speeds, showing the performance of the car as it keeps up pace with the jet in the sky, before the car accelerates rapidly, the sky is filled with stormy clouds creating and epic ambiance, 4k cinema style Super rough edit...but fun. This isn’t perfect...multiple continuity errors. But majority of the shots are from 1 image. QUICK THOUGHTS: + You can do a lot with one image. + 80% of the actual race is the same image + Cropped/zoomed in/isolated/etc + All the video tools offer something different + Runway: Object permanence + keyframes + KLING 1.6: Prompt coherence + action + Minimax: Creative motion + X factor  + Sora: Higher resolution + recut + This was my first shot at focusing on sound + It’s extremely hard to make things sound real + Much respect to sound engineers/designers + It’s super hard If you’re not familiar with MMAudio… It’s an open-sourced Video-to-Audio tool Available on Github (if you want to try) I’ll drop the link in the comments section. PS: (**No affiliation with BMW**) #runwayml #kling #minimax #runway #ai #innovation 

MIDJOURNEY MOODBOARDS → AWESOME. Broader Concepts = Cool Results. Odd thing is… If you treat moodboards…like a moodboard You get decent results. PROCESS: 01. Open Midjourney website 02. Select “Personalize” 03. Select “Create Moodboard” 04. Upload/Add images to MB 05. MJ will generate a “profile code” 05. Select “copy code” 06. Write a prompt + paste code 07. Generate HERE’S AN EXAMPLE: Profile Code: --profile ncwsycx BOARD FOCUS: + Theme: Organic Technology + Secondary: Futuristic 1980s + Textures: Gritty + Smooth + Color Contrast: Flat Primer + Neon + Colors: Radical Teal, Sapphire Crystal Blue, Purple Heat, Tangerine Orange EX PROMPT: interstellar mushroom --ar 4:5 --q 2 --profile ncwsycx --s 350 --v 6.1 QUICK THOUGHTS: + Broader image set = diverse outputs + Tighter image set = consistent outputs + You can use your own MJ images + Or you can source your own + For more unique outputs… + Try adding specific textures/colors/tones + Also try compositions/lighting/effects + The tighter your moodboard… + The less you need to text prompt + It’s easier to transfer aesthetic w/o over prompting Give the profile code a try…[--profile ncwsycx] LMK how it works. Going to explore the hell out of this feature…more coming. PS: Drew and I went deep on Midjourney : Fast Hours (LINK IN COMMENTS) #midjourney #runwayml #kling #minimax #ai #innovation 

MINIMAX OUTTAKES V6 → XMAS PARTY. When the drones land in New Jersey... Minimax never disappoints. It’s a good time to sit back…relax…and Just play with the tools…no agenda. >> This is purely for entertainment<< MIDJOURNEY PROMPT: flash photo, 1980s comedy film that was never made, and movie we never needed lol, Santa Claus and a stereotypical extra-terrestrial with an oversized head, big eyes, and a grey spindly body at a house party --p --ar 5:3 --v 6.1 MINIMAX PROMPT STRUCTURE: The camera [motion] as [subject + action] MINIMAX PROMPT EXS: 01. The camera follows as santa claus and the alien walk to pet a Reindeer 02. The camera follows as Will Smith enters the scene with a giant bowl of spaghetti, lol 03. The camera pans down revealing the Grinch laying on the ground looking up at the camera and waving 04. The camera pulls out as Santa walks over to kiss an attractive woman awkwardly QUICK THOUGHTS: + Minimax has a sense of humor + There’s a lot of errors in these videos + But it’s worth it + We can appreciate the absurdity + (Or at the very least, I appreciate it) + Hoping for more silly/funny Ai in 2025 + We all need a good laugh For everyone celebrating… Have an excellent holiday! #minimax #midjourney #runwayml #sora #ai #innovation

RUNWAY’S NEW FEATURE → Middle Frame. Another week…Another useful update. Last week they quietly released “Middle Frame” So now you can utilize (3) Keyframes In a single generation. PROCESS: 01. Open Runway GEN-3 02. Generate/Upload Images 03. Arrange Keyframes 04. Add a text prompt 05. Generate This is a solid feature…but It requires image consistency Between the Keyframes. KEYFRAME CREATION PROCESS: + Beanie Man → Midjourney + CREF + Land Rover → KREA LORA + NASCAR → Midjourney + Edit/Zoom + Me → Krea LORA MIDJOURNEY PROMPT EX: editorial photo, epic side profile shot of a 1980s NASCAR racecar, driving on a track, contrasting colors, spinning wheels, deep textures --ar 5:3 --p --s 200 --v 6.1 RUNWAY PROMPT EX: action footage of a Nascar racecar driving to the left, the camera zooms in on the racecar car speeding down the track at 30x speed QUICK THOUGHTS: + It’s a nice update…but it’s not a fix-all + Small details make a huge difference + And can also be a pain + More Keyframes = More Morphing + (If the images aren’t consistent)  + Focus on the small character/env. details + Might require some post-processing on KF’s + I find it works better for subtle movements + Huge motion is more difficult + The text prompt is still highly relevant + The keyframes need to match the prompt + For an easier keyframe workflow… + Try starting in closeup (ex Nascar) + And out-painting to expand the scene + The KF’s will stay consistent  + It will make the interpolation smoother More testing to come. Have Fun. PS: If you just getting started…there’s a Runway Cheatsheet in the comments section #runwayml #midjourney #sora #minimax #ai #innovation 

META AD WORKFLOW → KREA + SORA. Media buyers should like this… My friend Chris Erthel recently launched AdUniverse.ai (So I gave it test run) And did some mockups with Krea + Sora. WHAT IS IT: + Ad Creative tracking/inspo tool + You can look at META Ad spend + For some of the biggest brands + So you can build ad creative + Use data to make decisions. WHAT YOU CAN DO: + Analyze ad creative / landing pages + Track specific brands creative + Save ads to collections + Share boards with your team or agency MOCK AD CREATIVE PROCESS: 01. Krea: Train a model on your product 02. ADUniverse: Get inspiration 03. ChatGPT: Generate Prompts 04. Krea: Generate base images 05. Figma: Construct ads 06. Sora: Animate images QUICK THOUGHTS: + I have been using this product IRL + It’s super easy to use + Tracking brands is addicting and insightful + It’s awesome to see what they’re doing + And what trends are really popping + Super helpful with creating static ads + Using formats and templates that have worked + Estimated ad soend calc gives you an idea + Of what creative is really working well + However, it only tracks spend from the EU Disclaimer: + If you’re not familiar w. meta ads/paid media… + You need constant inspiration/iteration + It’s a volume game at times + And Ai can be your best friend here + We’re not looking for award winning campaigns + We’re looking for scroll stoppers + And creative that drives clicks PS: Link to AdUniverse in the comments PPS: Check out Chris if you haven’t already…awesome guy with a great mind. #midjourney #runwayml #sora #minimax #ai #innovation

MIDJOURNEY → RELEASES MOODBOARDS Probably my favorite feature this year. WHAT IS IT: + Create multiple personalization codes + By utilizing “Moodboards” + To create a “style” + (They can be super broad or specific) PROCESS: 01. Go to "Personalize" Tab 02. Select "Create Moodboard" 03. Add MJ Images or upload your own 04. MJ will provide a "profile code"  05. Generate MIDJOURNEY PROMPT STRUCTURE: high precision photo of [subject] --ar 5:3 —profile [code] --s 300 MIDJOURNEY PROMPTS: high precision photo of {Nike AirForce 1s, Octopus, Spiritual Healer, 1980s Firebird TransAm, Samurai, Orchid} --ar 5:3 —profile [code] --s 300 Profile Codes Used:  + Crystal Teal: --profile kwcr4qn + Burnt Orange: --profile dbcchkd + Harsh B&W: --profile m4xwcck + Neon Acid: --profile o563r8u QUICK THOUGHTS: + Consider these a mini-LORA + You can create your own “style code” + Create styles that are broad/versatile + And incorporate multiple styles/looks + Or you can create something specific + (like the images below) + This will make prompting way simpler + More focus on style curation than tokens + Makes mimicking your brand style easier + This is more about underlying functionality + Than just “curating” styles + This feature will probably be used for + Characters + Objects moving forward Really big fan of this update. Will be testing this way more. PS: Give these profile codes a test if you want to try them out #midjourney #runwayml #minimax #ai #innovation #chatgpt

When the New Jersey "Drones" reach Brooklyn... (With help from Sora + Suno) If this doesn’t make sense… Search “New Jersey Drone Sightings”  (That should clear it up) It’s Sunday…so let’s have some fun. >>This is purely for entertainment<< SORA PROMPT STRUCTURE: the camera focuses in on [subject + action + motion] in an [environment], [character details], [environment details]. [Emotion/Feeling]. [Film style]. SORA PROMPT: the camera focuses in on an alien with an oversized head and large eyes dancing rhythmically in an underground club, wearing a black hoodie, surrounded by humans with tattoos and trendy attire, who are losing themselves in the music. Showing the intermingling of humans and aliens having fun. The pulsating sounds of techno move the crowd in unison. high contrast black and white archival footage. super crisp slow motion SUNO PROMPT: Deep, bouncy hypnotic techhouse. bongo stabs, crisp hats, vocal chant loop sample, 124 bpm QUICK THOUGHTS: + Text-2-Video in Sora needs details + It can cut together multiple shots  + In one generation + This prompt has worked multiple times + Without many failures + Meaning, the structure probably translates  + Will test this one more Have a weekend. #runwayml #minimax #kling #midjourney #ai #innovation 

SORA → 1 VIDEO...MULTIPLE OUTCOMES. Remix is the most consistent feature (imo) Use for big/small changes. Super simple. PROCESS: 01. Generate a video 02. Select Remix 03. Select Sublte/Mild/Strong Remix 04. Change/add/remove what you want 05. Hit Submit ORIGINAL PROMPT: A cinematic behind-the-back following shot of a motorcycle in a hazmat suit, driving away from us through a thick bail jungle dirt road, thick fog with hazy light after a rain storm REMIX PROMPTS: 01. Change motorcycle to an ATV 02. Change motorcycle to a Bulldozer 03. Change motorcycle to a Hovercraft 04. Change motorcycle to a jetski, change the environment to a beautiful atoll ocean 05. Change jetski to snow mobile, and change the environment to the swiss alps with snowy peaks and bright crystal blue sky QUICK THOUGHTS: + Not all will be perfect + But you can keep remixing  + This is Sora’s least glitchy function + Sometimes you need to run the same video + Through 2-3 iterations for big changes + The exs. Below are larger changes + But you can make smaller changes also + Ex (make the road less muddy) + But choosing between subtle/mild/strong + Is a nice balance for this feature Continuing down the rabbit hole all weekend. Have fun. #minimax #runwayml #sora #midjourney #ai #innovation 

SORA → BLEND + REMIX This is Sora’s differentiator (IMO) And using them together is cool. But what do they do: BLEND: + Mix two videos together + Think keyframes with full video + Blend strength/weight is dynamic REMIX: + Change small or big details + Upscale or extend + Also can smooth out errors PROCESS: 01. Generate 2 separate videos  02. Use Blend to mix them together 03. Use Remix to fine tune SORA PROMPT 1: A cinematic front view tracking shot of a motorcycle in a hazmat suit, driving away from us through a thick bail jungle dirt road SORA PROMPT 2: A cinematic behind-the-back following shot of a motorcycle in a hazmat suit, driving away from us through a thick bail jungle dirt road, thick fog with hazy light after a rain storm SORA REMIX PROMPT:  The biker makes a skidding turn, wearing a yellow hazmat suit QUICK THOUGHTS: + There were multiple runs on this + Mostly out of curiosity + There’s infinite potential with this + And it makes Sora’s potential very creative + Think of it like image ref in Midjourney + And keyframes in Runway/Luma/Kling + But with full video instead of images + It can help gen more complex shots + So think about pieces of a whole + Not just one shot gens These are not perfect…and still need work. But you can see how to mash ideas/concepts. Have fun. #sora #runwayml #minimax #midjourney #ai #innovation