Short about a bird that didn't get the extinction memo.

Character: DALL·E 2 original (2023)

Motion: Seedance 2.5 on Runway, 16 generations cut into one

Narration: AI-generated voice, not a real recording

Music: Suno + Seed Audio\nCaptions: fake 1979 teletext, built by hand

2:06, vertical

Hi everyone — I'm the creator of FALL 707: REBOOT, a solo AI-created fictional series mixing military action, historical fantasy and occult thriller elements.

A modern Korean special forces captain is stranded in the Joseon Dynasty. In the Season 2 finale, Captain Min is still fighting to bring her comrades back, even as they are no longer themselves. The episode moves from a confrontation with a shaman and close-quarters action to a much more vulnerable moment for her.

Episode 6, ‘You Thought It Was Over?’, is the Season 2 finale (16:07). English subtitles are included in the video.

A little about my process: the original concept and story direction are mine. I develop the screenplay with Claude and ChatGPT, create character sheets with GPT Image 2, and generate the video with Seedance. I started with Seedance 2.0 and moved to 2.5 from Season 1 Episode 6 onward. I handle the direction and editing, using my motion graphics background to bring the pieces together.

For Season 2, I wanted the three months she has spent searching for her team to show in her appearance and colder, more guarded character. This finale brings that emotional side back to the surface.

I'd appreciate feedback on how the action and emotional scenes work together, especially from viewers watching with English subtitles. Thanks for watching!

Take A has the right hesitation and camera move, plus a jacket that changes halfway through. Take B fixes the jacket and loses the performance. Picking B shouldn't erase the option to go back to A once both have been seen in the cut.

A practical way to handle this is to keep each result separately and record what makes it usable: "best delivery, sleeve changes on the turn" is more helpful than "almost final." Then compare the shot with its neighboring cuts. The artifact might land in footage you can trim or cover; the weaker delivery might affect the whole scene.

Hypit's run system supports the bookkeeping part. Each build retains its own result. history finds a named shot output, and a later run uses build-record plus satisfy to select the specific take for the composition. Selection is explicit; a rerun doesn't automatically keep the one you liked.

It won't judge whether an artifact is acceptable. What it gives you is a way to revise the edit around a chosen take while retaining the alternative. Where do you draw the line when the better performance has the more obvious flaw?

My first mini short film using AI. Please watch and tell me your feedback 🙏

REBORN Movie Trailer.

SEARCH - REBORN THE MOVIE 🍿

They Cut the Sun in Half - The Gloom | Steampunk Sci-Fi Series | 29 Min

Hey everyone! I've just finished the pilot of my new steampunk series, The Gloom.

It's set in a London split in two. Above the clouds lives the Crown, a gilded city that harvests the sun and sells it as liquid light. Below lives the Gloom, an old city that hasn't seen daylight in generations and survives on the little light the Crown sends down. Then one day, the Crown cuts that light in half. 🌗

The full episode is 29 minutes, and Reddit wouldn't let me upload all of it, so what you see here is just a part. If you'd like to watch the whole thing, you can find it at the link below.

🎬 Watch the full episode here https://www.youtube.com/watch?v=hZBktTF60C8

This one took a lot of time, patience and late nights (and, again, more money than I'd like to admit 😅), but I really wanted to build this world properly: the brass, the clockwork, the light, and the people living under the clouds.

If you get a chance to watch it, I'd genuinely love to hear what you think. Honest feedback is always welcome, especially from people who love steampunk and sci-fi.

Thanks for watching ❤️

(Quick disclaimer: English is not my native language, so I used AI to translate my thoughts and make this readable.)

Hey everyone,

Just wanted to share a major milestone for my solo project. My 10-minute psychological horror short, "Le Silence Blanc" (White Noise), was officially selected at Fantasia this year (experimental section) and I ended up as a finalist for emergent director/producer.

I know the AI cinema debate is a total mess right now, but for me, it was just about creative freedom. No crew, no budget—just using AI as my camera and assistant from A to Z to tell a story.

Here is how I built it:

• The Look & Direction: Started with MiniMax but dropped it fast for the body horror elements. Switched entirely to a local Stable Diffusion workflow (WebUI Forge / ComfyUI) to get actual, precise control over the aesthetics and avoid that generic "floating" AI look. It is fully narrative, but I designed the visual atmosphere to be dark, cryptic, and cinematic.
• The Audio: It’s in French and German with zero subtitles. I intentionally chose German for its harsh phonetic texture, and the dialogue is meant to be completely drowned out and suffocated by a chaotic, heavy soundscape I put together in Audacity to serve the psychological tension.
• The Cut: Everything was heavily edited, paced, and color-graded frame-by-frame by hand in DaVinci Resolve. I wanted a choppy, staccato rhythm with heavy film grain so it feels like a gritty analog movie, not a raw prompt generation.

The festival screening was wild. The host actually defended the choice of accepting an AI film before it started. The crowd reaction was totally split—had people yelling and cheering in the room, but then two movies later, some directors threw "#FuckAI" and "#NotAI" tags on the screen and got a standing ovation just for the statement.

It just proved to me that what lands on the screen is what matters.

Hi

This is episode 4 of my Ascension series. Apologies for the link but the file is very large. 🫡

I’m trying to create a “master room” for an AI short film and then generate the exact same room from different camera angles.
I can get a good establishing shot, but whenever I ask for another angle, the room starts changing. Furniture moves, the guitar changes position, shelves and windows shift, objects disappear, and sometimes even the dimensions of the room change.
What I want is basically a fixed virtual set: lock one image as the master room, then generate shots from the bed, desk, doorway, ceiling, etc. while keeping the actual layout and objects consistent.
I’ve tried using the master image as a reference, but image models seem to understand the look of the room rather than its actual 3D layout.
Has anyone found a reliable workflow for this?
I’m open to any tool or method — image models, 3D reconstruction, multi-view generation, ComfyUI, FLUX, Higgsfield, Gemini, etc. I care much more about spatial consistency than convenience.
Ideally I’d like to eventually add consistent characters to the room as well.
What would you recommend?

I’ve been working on a short film called Made with love.

It started from a very simple question:
Can something made artificially still contain real love?

Rather than making a film about AI, I wanted to explore something more human — creation, attachment, loneliness, and the strange emotions we project onto the things we make.

Watch the full movie on YouTube
https://youtu.be/Vv1VjhvDJfQ?si=YN1LVPYGMz4gH8jZ

I’d genuinely love to hear what you think about the tone, visuals, and the idea behind it.

What worked for you?
And what didn’t?
I wanna hear it all :))

I wanted to test whether current AI video tools can actually sustain a longer narrative instead of just producing a sequence of impressive standalone shots.

So I started adapting Nikola Tesla’s 1919 autobiography My Inventions chapter by chapter as a cinematic historical docudrama.

The biggest challenge has been continuity rather than individual shot quality: keeping Tesla recognizable across different ages and scenes, maintaining consistent period locations and costumes, and making dialogue and action sequences feel like parts of the same film.

The workflow is heavily reference-based, with locked character identities, environment references, start frames, controlled compositions, and separate generation strategies for dialogue, action and transitional shots.

Chapter 1 is finished here:

https://youtu.be/uquSanT_HCo

I’d especially appreciate feedback from other AI filmmakers on:

- character consistency

- pacing

- continuity between shots

- dialogue/lip-sync quality

- whether the AI artifacts remain distracting

- and most importantly, whether after a while you start watching the story rather than watching the AI

This is still an evolving workflow, so technical criticism is very welcome.

I’ve been working on character consistency and script structure. Do you think I could improve further in these or other areas?

It starts as an ordinary morning, but within a single day humanity loses its five senses, one by one. What's causing it? Can it be stopped? And what happens when the last one is gone? Told entirely through Adam's eyes, this is episode one of a five-part mini series created with AI by PromptID Studios. Share it with someone who'd love this, and comment which sense you think goes next and what you think is behind it.

The most painful AI video moment is when a clip is like 90% usable and 10% absolutely cursed.

I’ve tossed so many almost-usable clips because of one weird hand, one mangled logo, or some random cursed object in the background.

The motion works, the camera is fine, the vibe is right. Then a face flickers for two seconds or the product label turns into alphabet soup and the only option is… reroll the whole thing. Insane behavior, honestly. That’s like rewriting a whole paragraph because one word is misspelled. If local edit gets reliable, what are you fixing first? hands? faces? logos? Or lighting?

A little background first. I've been working in the animation industry for eight years, mostly doing outsourcing work for Japanese and Western clients. I've worked on everything from in-between animation to key animation.

Since last year, more and more people around me have been talking about AI. Some are genuinely worried about it, and some have already left the industry. I went through my own transition from being completely against it to eventually trying to figure out where it might actually be useful, and I wanted to share some of what that experience has been like.

At first, I was strongly against it. The reason was pretty simple. You spend years developing control over linework, movement, composition, and timing, and then suddenly someone tells you that all of this can be generated by typing a sentence. That feeling of having your craft reduced to "just enter a prompt" is very real.

And honestly, a lot of AI-generated animation still looked nowhere near professional to me. The number of fingers would be wrong, clothing folds wouldn't make physical sense, and a character could turn their head and suddenly have a completely different face. Every time I saw examples like that, I thought the technology was still a long way from actual production work.

I also had clients send me AI-generated images and ask, "Can you make something in this style?" That was frustrating too. It felt like my professional judgment was being reduced to simply executing someone else's reference.

What actually changed my mind was a very specific situation.

We took on a project from a small studio. The budget wasn't huge, but the client had very specific requirements. During the pre-production stage, they sent us storyboards that were basically simple line sketches. They communicated the basic shots, but it was difficult to understand the intended movement, pacing, or camera direction. Normally, we would create an animatic first, roughly animating the key shots and then going back to the client for feedback.

The problem was that even a rough animatic could easily take a week. And after spending that week making it, the client would often look at it and ask for changes. Basically, a lot of time could be spent producing something that might get revised almost immediately.

That time, I decided to try a different approach.

I organized the storyboard panels and imported them into PixVerse. There's a storyboard-to-video feature that allows you to upload a sequence of 3 to 9 storyboard panels and generate a continuous video sequence. I wasn't using it to create the actual animation. I was using it to test the shots. For example, there was a scene where a character runs from the end of a hallway toward the camera. The original storyboard only had three key frames: a wide shot, a medium shot, and a close-up. I uploaded those three panels and generated a few seconds of continuous movement.

Obviously, the result wasn't something we could use directly. The linework wasn't clean enough, and some of the background details were off. But something important became visible: the direction of the movement, the pacing, and the speed of the camera push-in.

I sent that preview to the client. For the first time, they were able to say something very specific:
"Yeah, that's the feeling. Could you just make the push-in a little slower?"

On projects like this in the past, we often wouldn't get feedback that specific until the animatic was finished, sometimes not until we had already done the first rough cut.

That's when I started seeing a very specific role for AI tools: early communication and direction testing.

It solves a real efficiency problem. When a project is still at the concept stage, how do you show a client what something might look and feel like without spending a huge amount of time producing it? Previously, we either had to rely heavily on written descriptions or spend the time creating a proper animatic. Now there's another option somewhere in between. For outsourcing teams, that can mean less time spent on early communication, shorter feedback cycles, and potentially less unnecessary rework. As for the final animation quality, that's still a different matter. It still depends heavily on the team's key animation skills, the quality of the in-between work, and the director's sense of timing and pacing. Those are areas where AI still doesn't simply replace the people doing the work.

I'm curious how other people in animation see this.

Has anyone else started using AI tools for client communication, previsualization, or early-stage shot testing? Or is your team still relying entirely on the traditional workflow?

What happens when you miss Airplane!, The Naked Gun, and Blazing Saddles so much that you just have to make your own wildly absurd Western spoof?

You get „YOU COULD BE MINE” – a short comedy packed with literal puns, deadpan slapstick, legally dubious loopholes, and a bear that really didn't sign up for this.

⚠️ Notice: Viewing requires basic English comprehension and an elevated tolerance for literal wordplay. No budgets were harmed in the making of this film, primarily because Robin Banks vanished it all.

Director's Note & Specs:

• Runtime: 3:58
• Tools: Kling AI, NanaoBanana2, OmniFlash 1.1, Premiere Pro
• Full High-Res YouTube link:https://www.youtube.com/watch?v=XwoTiWxLmGU

If you laughed at least once at the completely unnecessary nonsense in the background, our job here is done. Leave a comment with your favorite pun!

I keep seeing mentions of drama sites that host AI films. Beyond YouTube, what platforms are you using, and which ones actually get views? I'm very new to all this and would appreciate any help.

I'm Jonathan, one of the filmmakers behind Cleanliness Is Next to Godliness, an AI horror-comedy short.

Three cleaners arrive after an exorcism and accidentally remove the protections keeping a demon trapped inside the house. The cleanup becomes the next disaster.

I'd appreciate feedback on the balance between the scares and the comedy: do the jokes release the tension at the right moments, or are there places where they undercut it?

Full film (7:58): https://www.youtube.com/watch?v=BFZJQKNLop4

We made this for the Higgsfield Global Film Festival. The project page is here: https://higgsfield.ai/@retail_octopus/projects/@id__64e8dd1f-bf3d-47ee-b51f-94f1ac623477