Flesh Syntax1.5 is here and I wanted to give it a quick review.
Still getting a hold of this youtube thing and figuring out what this channel is supposed to be.
i'm a conventional filmmaker working in the film industry and working with GenAI, leading teams to push the boundaries of what AI video can do in a production setting.
Overall, I feel like Cinema Studio is the most accessible, pro-sumer tool out there currently that attempts to cater to serious, cinematic filmmaking / indie filmmaking projects. But there's still a ways go to!
Real Filmmaker Reviews Higgsfield Cinema Studio (NEW UPDATE)Flesh Syntax2026-01-10 | 1.5 is here and I wanted to give it a quick review.
Still getting a hold of this youtube thing and figuring out what this channel is supposed to be.
i'm a conventional filmmaker working in the film industry and working with GenAI, leading teams to push the boundaries of what AI video can do in a production setting.
Overall, I feel like Cinema Studio is the most accessible, pro-sumer tool out there currently that attempts to cater to serious, cinematic filmmaking / indie filmmaking projects. But there's still a ways go to!
Love it or hate it? Lmk your thoughts.Google Flow 2 Test Scene – PistachioFlesh Syntax2026-02-26 | Google Flow's revamp just dropped, so I thought it was a good time to test it out.
Veo 3.1 has improved its audio outputs significantly! And the interface is, well, definitely a step up from what it was before, though dragging and dropping things into a narrow bar instead of a drop box was a bit finicky for me. Learning the new controls and where they now live will take some getting used to.
This short scene was inspired by the Google Flow Discord's daily prompt - which happened to be "pistachio"
The unofficial mascots for the discord are an otter and a snail for some reason, so I thought it was a good opportunity to do some light branding placement type stuff, which is something I do a lot for my work. Used Nano Banana Pro thru Cinema Studio, along with Photoshop to produce the images and animated in Flow.Higgsfield Cinema Studio 3D: How It Really Works (+ Fix )Flesh Syntax2026-02-23 | Today we're doing a deep dive into @HiggsfieldAI Cinema Studio 2.0's new 3D Scene tool, which I feel has a LOT of potential, but it seems like a lot of folks gave up on it after their first failed experiments.
I was also frankly, feeling a bit disappointed with the tool ***UNTIL*** I figured out the mechanism behind it!
With a few tweaks, you can have maximum control over the image, customize references, and achieve creative shots with depth, likeness and detail.
This video took waaay too long to make, and so much ended up on the cutting room floor for time!
I started out wanting to just do a deep dive on 3D scene. I had all this stuff where I explained how Gaussian Splats work etc. But then I realized it might be more useful to just explain to folks how to actually get the most out of this tool and make these outputs work properly.
Because at the end of the day, all the outputs need is a nudge via prompting and specificity, and you can really play around with this tool and get cool results!
Meanwhile, Higgsfield was actively updating the tool every day as I recorded....
When I started the vid, the outputs were not named, then they were called "Cinema Studio 3D Image Recompostion" (which I felt was most accurate) but as of today, the outputs are called "3D Image Shot"
But I can understand why they have had a hard time naming these things because they don't work how most people might assume a "3D tool" would actually work - which I explain in detail in the vid.
I know this vid is pretty long - but I think there's a lot of valuable info in here not just for using 3D Scene, but also for prompting things on your own. The prompting method I explain in the video - which I think of as Directional / Spatial prompting - and yeah, I also realize that the prompt keywords used is not groundbreaking and likely terms that most AI folks are using already. It's more that the simple structure of the prompt and "location-based" nature of the prompt is what gives it its power. I've found it's the most effective prompt method, and coupled with a sketch / image guide it's pretty unstoppable.
Anyhow, I'm hoping that my next upload will be my deep dive on that "secret" method - where I'll walk through examples in more detail and explain more thoroughly how I get one-shot / one batch iterations.
It's a method that I use on my professional projects, and the rate at which I get one-shots is higher than any other method I've used, to the point where if I don't get a perfect output, I'm surprised (or at the very least, I can easily identify the failure point and get the perfect output quickly).
Chapters: 00:00 — Intro 00:36 — What Cinema Studio is 00:58 — 3D Scene overview: 2D image → 3D → snapshots → renders 02:09 — Successful renders: snapshot vs final output 02:36 — Where it goes wrong: likeness drift, angle errors, weird glitches 03:41 — Core theory 05:03 — My "Secret" workflow - PROTIP: reference-stacking 06:32 — Side-by-side test 07:56 — Why abstract snapshots / simple sketches still work: contrast, shapes, and color anchors 08:45 — Using Whisk to learn how AI sees 11:41 — Rebuilding the same scene with SGG: rescued compositions! 13:00 — The hidden prompt bug: why specificity changes the result 13:55 — Screenshot-able prompt keywords for new prompt style 14:06 — The FIX: how to recreate the shot and fix flaws (workflow) 14:13 — "Secret" workflow Step-by-Step: Recreate → Image → add original reference → describe shot 15:29 — Before/after comparisons: same snapshot, better prompt, usable output 16:58 — Wrap: why 3D Scene is useful and fun - don't give up!
I know that a lot of GenAI releases are centered around speed, but when it comes to creativity, process and experimentation are just as - if not more - valuable. It's in the experiments that we learn the most and hone our taste and craft.
I think 3D Scene is a really interesting space to experiment with.
But when I saw this hit the market, I HAD to take a look.
Google just released Genie 3 - a new AI tool that generates 3D worlds in real time using gaussian splat technology.
In this video, I jump in cold and explore what it actually feels like to use this tool - from the weird to the wonderful. From 1st person / 3rd person movement, the strange technical outputs, surprising movement from air, land, sea, and space.
This is NOT a polished demo and not a sales pitch either this is just a cut-down of my real first impression experience experimenting with this tool from a filmmaker’s perspective.
I know this tool is way out of budget for many people. While I think it's super fun and has potential to be... more fun?
At this time, i can't say it makes much sense to me for anyone to drop 200+ just to try it out unless you're REALLY into video games. The current low output quality, the glitches, and vast limitations of the tool make it hard to pinpoint a use case in its current state, in my opinion.
If you have a world you want me to try building - leave a comment and if I get a chance to I'll see if I can iterate it!
00:00 – Preview 00:08 – First Attempt + 1st Person Character??? 00:26 – 1st World generation - First Impressions! 00:33 – 1st person... or 3rd person...? Existential crisis 00:55 – 59-second limit + output specs (WEIRD) 01:17 – Gaussian splats / real-time rendering explained 01:38 – Odd glitch - automatic movement vs camera control 01:53 – Underwater World! 02:34 – Strange glitches 02:49 – Running in Outer Space World! 03:17 – Movement rules 03:31 – Tiger in a Mansion World! / Google Flow memory & iteration bleed 04:03 – Vehicle test, speed, failsafe 04:25 – Filmmaking utility? vs game utility? 04:58 – Final thoughts & where this could goHow I Use ShotDeck for Cinematic AI Prompts (Filmmakers Secret)Flesh Syntax2026-01-12 | ***LINK: shotdeck.com/welcome/home
PROMPT (I've changed a couple of angled brackets to slashes because of youtube restrictions):
Tags: sleuth, small apartment bathroom, looking directly up, redhead, big hair, adjusting her sunglasses, casual, light yellow t-shirt, curious, cinematic, sideways Genre: Comedy, Crime, Mystery Time Period: 2020s Frame Size: Wide Shot Type: Clean single, Directly Overhead Composition: Center Lighting: Soft light, Top light Lighting Type: Daylight, Sunny Time of Day: Day Interior/Exterior: Interior Set: Funeral Home / Bathroom Story Location: ... Florida Yulee Color: Balanced
On my last video I noticed a couple of comments about how I used ShotDeck metadata as a base for my prompt.
Wanted to do a walkthrough where i demonstrated my process live with a brand new shot, to give ya'll a sense of how i go about building out the data and optimizing it for GenAI.
Someone also asked about LLM's / automated prompts - short answer is, I do not use LLM's for this process, as i've had a lot of success with manually building out the prompts.
About 2-3 months ago, I was using LLM's a lot for prompting, but recently I barely use that method at all, and, outside of this metadata extraction method, I usually prefer to write my prompts from scratch.
Though I'm sure that someone could create an LLM workflow with ShotDeck that works for them :)
**Also not sure if it needs to be said (still learning youtube) but this is NOT a sponsored video or anything. Just sharing something I use as a shortcut for making cinematic AI film and video prompts!**Grief.Flesh Syntax2025-12-20 | Was thinking about themes of grief tonight.
after my Mom died, I worried I'd never see her in my dreams. Finally when I did, I wanted to stay asleep - stay in that dream with her, just a little longer.
I wondered if my Dad ever dreamed about her too - and if he felt the same.
@HiggsfieldAI Cinema Studio for image gen Nano Banana and Photoshop for editing Higgsfield Cinema Studio, Higgsfield Lite, Seedance for video Higgsfield Cinema Studio native audio and Epidemic Sound for sound effects and ambient soundtrack.
AI mircoshort film.
#aifilm #aifilmmaking #aishorts #aiMicroshortGoogle Veos Camera Motion Feature - All Options DemoFlesh Syntax2025-11-08 | I tested all of @Google Veo's new Camera Motion options so you can see clearly what it does.
0:00 - Original Clip 0:08 - Orbit Right 0:16 - Orbit Left 0:24 - Orbit Up 0:32 - Orbit Down 0:40 - Dolly Out Zoom In 0:48 - Dolly Out 0:56 - Dolly In Zoom Out 1:04 - Dolly In
Things i noticed:
- Using this feature slightly degrades the footage, similar to "extend" but not quite as bad. - the camera movements are simulations that appear to have more human programming behind them than data. I say this because they are pretty inaccurate compared to how they would look if actually shot IRL, and feel more like approximations a VFX artist might make if they had to comp together a similar looking shot using 3D software (backgrounds feel like wallpaper). - despite that, i was impressed that it adhered to the selected camera motion without issue. Each iteration shown is a first try, except for orbit up - i had to redo that one because her foot got all wonky. - dolly zoom is cool, but I wish they gave more normal camera movement options that we'd actually use. dolly zoom rarely has any purpose in a bigger film or video project. It's more like a meme or trend. For example, dolly left, dolly right, truck up / down, or, crucially, tilt! Like, very rarely will projects call for all this fancy stuff. - Small background elements get glitched or warped (specifically elements not visible in first frame of OG clip)
I MIGHT make a video where i briefly talk about this stuff in more detail - just wanted to get this up now so people can see. I wanted to demo a clip that had deep space and lots of foreground / background / midground elements to really see the moves proper.
#veo #aifilmmaking #aifilms #aifilm #googleveo #googleflowBreakfast - AI Horror Film | Halloween ShortFlesh Syntax2025-10-31 | Happy Halloween!!!
I created this for the @curiousrefuge Third Annual #AI Horror Film Competition
This thing went through SO MANY versions!!! I learned a ton while creating it. There's a ton of stuff I'd like to change or update or adjust, but unfortunately, given the fact that Whisk and Veo is constantly in flux with updates every couple of days at this point, I feel like I probably have to lay this to rest.
Wanted to share a new version today - but due to the most recent update (yesterday or the day before) I'm unable to finish my work on the scene I was adjusting :/
Biggest lesson I learned out of this project is that AI short films at this point must be made quickly, because you never know what update will happen and completely disrupt the mood or flow of your project.
For example, after one update, all of the characters were singing their lines, and it was a total nightmare trying to figure out the workaround for that.
I might make a video walking through the challenges of the film - definitely will approach my next one differently to improve consistency, pacing, and audio etc... But wanted to share regardless.
Best of luck to the other candidates in the contest! Some seriously good films this year and i'm sure the judging will be difficult.
#aihorrorshorts #aifilmmaking #aifilm #halloweenai #halloween2025Urban LegendsFlesh Syntax2025-09-12 | There are 7 ghosts hidden throughout this video. (13 ghosts total* ) Do you see them? * Short horror film based on a "found footage" film I shot a number of years ago.
After my Mom's death, I have been reflecting on themes of mortality, risk, living in the moment and the exhilaration of youth. Mom was always into spooky stuff (she was into true crime and horror video games like Doom, Resident Evil, and Silent Hill... I was too scared to watch her play Silent Hill as a kid!)
She hated fantasy though. If anything was too "unrealistic" or "fantastical" then she would immediately switch channels. The ghosts had to feel real.
So I found myself making this film in her honor, with the goal of creating something that felt gritty and "real."
The irony of course, is that I made this with AI.
The video was made with Google Veo, and the audio is a mix of Suno AI and Google Veo 3 - Quality audio.
#googleveo3 #googleveo #urbanfolklore #urbanlegend #halloween #scarystories #aifilm #aifilms #aihorror #aihorrorshorts #aihorrorfilm #horrorstories #horrorfilm #horrorshorts #halloween #halloween2025 #halloweenshorts #halloweenAI #foundfootage #foundfootagestyle #blackandwhiteAIGhost WavesFlesh Syntax2025-09-06 | I'm a visual artist and filmmaker by profession. After my Mom's recent death, I started getting work related to AI, and eventually got into Suno. I've found that making songs has been an amazing creative outlet for me to process themes of spirituality, technology, and human connection during grief.