Pinned Tweet
I've always been obsessed with AI films, not as a trend, but as the kind you spend hours building. The ones almost no one sees yet, and you keep going anyway because you care that much.
That's why I've been building Fluttercube in the background. A place to watch and share AI-made films, series, shorts and experiments, free at full quality, where anyone can post.
It's live → fluttercube.com
If you know someone who makes AI films, or just loves watching them, please share this with them. Let's build this community from the very first film.
2
98
The full film is up on fluttercube.com. I built the site because I couldn't find a good place to share AI films or watch other people's. Come take a look.
🤖 Made with AI
1
74
Hopping between tools slowing you down? Many creators are now struggling with the complexity of handling multiple applications. What tool has been your biggest time-saver recently?
#CreatorStruggles
#AIFilmmaking
51
One AI video project. Six stages. Six different tools. That's not a workflow -- it's a scavenger hunt. No wonder tool-hopping fatigue is now the #1 complaint louder than any quality issue.
My tech stack includes Claude Code with skills, Flow, Gemini, Image 2.
What's yours?
#AIFilmmaking #VideoProduction
50
Exciting news! Kling 3.0 introduced 4K native video support, 15-sec clips, and multi-shot storytelling. This pushes the creative boundaries in AI video production! How will this impact your workflow? #Kling #VideoProduction
52
The biggest bottleneck in AI filmmaking isn't generation quality anymore.
It's the 5-10 tool handoffs per project: generate → upscale → edit → grade → sound → export.
What's your current stack?
#AIFilmmaking
1
26
Every week the AI filmmaking debate gets framed as replacement vs. no AI.
But the actual community is mostly people augmenting tiny teams or working solo.
The loudest voices in the debate aren't the ones actually using the tools daily.
Who resonates with this?
1
21
Pro tip: AI tools make clips, not films.
Editing workflow is the key challenge, not just generation quality. How do you ensure consistent characters across scenes for seamless storytelling?
#AIfilmmaking
1
24
Runway launched "Characters". Is this something useful for AI Filmmaking? Anybody here tried it?
#AIFilmmaking
51
OMG, tried Flow's built-in 'Dolly in' (Veo 3.1). What a disaster :-).
Make no sense. Am I missing something? Does anyone has a good result with it?
49
AI filmmaking is moving fast, and honestly, not everyone's feeling great about it. C2PA verification is still pretty rough around the edges. Hollywood is pushing back hard. And a lot of creators are just not sure they trust it yet.
Curious what you think, what's the biggest thing hurting AI filmmaking's credibility right now?
41
We’re having a calm, snowy winter night today.
So I created something that reflects this mood.
Honestly, how crazy is it that you can create a movie based purely on how you feel?
Story to be continued — stay tuned.
Tech stack:
• Google Gemini for prompting
• Google Nano Banana Pro for master reference and scenery images
• Flow for video generation (Veo 3.1 – Fast): Ingredients to Video and Frame to Video, stitched together in Scene Builder
• Flow’s built-in image editing for future adjustments
Tips:
• Expect multiple runs — the same prompt + same ingredients still produce different results
• Save a frame as an asset, refine or change it with built-in Nano Banana, then reuse it for a new generation
• If Veo 3.1 comes with an interesting object in the generation, cut and convert it into a Master Reference image without the background.
#VEO3 #Aifilmmaking
167
Tested Veo 3.1 scene extension by adding narrative beats to a clip.
Started with: Celtic warrior walking on a cliff.
Extended to: She stops, whistles, a white wolf joins her, camera pulls back to an epic wide shot.
What worked:
– Identity held across the entire extension (face, costume, hair)
– Wolf integrated naturally, with matched lighting and style
– Camera pull-back felt cinematic, not stitched
– Wind physics stayed consistent (hair, sash, wolf fur)
What didn’t:
– Audio wasn’t extended; it regenerated from scratch
– Audible transitions between clips
– Sword disappeared in the second shot (hidden behind the wolf in the last frame 🙂)
Takeaway:
Scene extension nails visuals. Audio is not a seamless handoff, expect to redo sound design if your source clip has audio.
Tip: If an element must appear in the next shot, make sure it’s clearly visible in the last frame.
For now: extend picture, rebuild audio separately.
Anyone else testing this? How are you handling the audio gap?
#Veo3 #AIfilmmaking #AIvideo
57
Tested Veo 3.1 4K upscaling on a close-up.
720p → 4K. Same clip, side by side.
What I noticed:
- Freckles are slightly more defined
- Hair strands are a bit sharper
- Skin texture gains subtle detail
- No artifacts or weirdness introduced
Honest take: It's incremental, not transformative.
The upscaler is competent, it doesn't break anything. But I wasn't blown away. It feels like a clean upscale, not AI "enhancement" that adds detail that wasn't there.
Maybe I'm missing something. Here are both versions—720p and 4K.
What do you see? Worth the extra render time, or skip it?
#Veo3 #AIfilmmaking #AIvideo
Tested Veo 3.1's Ingredients feature for multi-shot consistency.
The question: Can you build a sequence: wide, medium, close-up, and keep the same character across all shots?
Setup:
- Slot 1: Portrait (locks face)
- Slot 2: Full body in costume (Claude Skill for outfit prompt, then, Nano Banana Pro + Master Reference for image of character in outfit).
- Slot 3: Environment reference (Nano Banana Pro)
Generated 3 shots of a Celtic warrior on a stormy cliff.
What held:
- Face: Bone structure, freckles, eye color. The same person in the close-up as in the wide shot.
- Hair: Red with braids, wind physics consistent across all three.
- Costume: Tartan sash, leather armor spirals, Celtic embroidery - all tracked through motion.
- Environment: Lighting and color grade matched without manual grading.
What this means:
Ingredients work and isn't just for single clips. It's a shot-matching system.
You can build a sequence with wide/medium/close-up and cut them together without identity drift.
Tip: Use a portrait AND a full-body reference.
Slot 1 locks the face, Slot 2 locks the costume. #Veo3 merges them.
#Veo3 #AIfilmmaking #AIvideo
1
193
Tested Veo 3.1's native 9:16 vs. cropping 16:9 to vertical.
Same prompt, same scene: woman reaching up to catch snow in a forest.
The differences:
Native 9:16:
- AI composed for vertical. Trees tower overhead, full body fits boots-to-fingertips, action stays in frame.
- Full pixel density. Snowflakes stay sharp, coat texture crisp.
Cropped 16:9 → 9:16:
- Arms kept exiting the sides. The AI was "thinking horizontally", so the motion wasn't contained.
- Cropping discards ~68% of horizontal pixels. The center gets blown up, snowflakes go soft.
The real difference isn't just resolution - it's that the AI
composes differently when it knows the canvas is vertical.
If you're making content for Reels/TikTok/Shorts, generate native 9:16. Don't crop in post.
#Veo3 #AIfilmmaking #AIvideo
1
56
Prompt:
Static medium wide shot of a woman in her early 30s with shoulder-length dark brown hair, wearing a long burgundy wool coat with black buttons, reaching upward with both hands to catch falling snow, face tilted toward overcast sky, standing in a snow-covered forest clearing. Full body visible from boots to raised fingertips. Soft snow falling around her, pine trees framing the edges. Shallow depth of field, cool desaturated color grade, locked-off camera, 35mm film grain, joyful contemplative mood.
28





