AI video artist. The goal is for you to not be able to tell.

Joined May 2026
BEYOND THE PATH (その道の先) Episode 1: The Boy Who Cut the Sky #Anime #AIAnimation #AIFilmmaking #Seedance
5
2
2
17
1,074
SAVE YOUR TOKENS. If you're using AI to make Blender previs for Seedance 2.5, don't overcomplicate it. Same prompt. Two previs approaches: 1. Barebones blockout 2. More detail + animation Result: barebones won. The more detail you put into the previs, the more literally Seedance follows it. So keep it simple: Block the shot. Keep detail and animation to a minimum. Let Seedance fill out the rest. #AIAnimation #AIFilmmaking #Seedance #Astra
15
45
2
324
21,383
Prompt: Exact same style as @Image1. 30-second dynamic shot, one continuous take. Use @Video1 for overall character blocking, action timing, movement paths, spatial relationships and camera motion. The knight, dragon, mountains and buildings in @Video1 are previs references for staging and scale; rebuild their appearance completely in @Image1’s exact style. The camera begins in a tight close-up of @Image1 standing at the edge of a towering mountain cliff as winds whip his blond hair and billow his burgundy cape over his ornate silver armour. Hold for one second, then slowly widen to a close-medium shot as he leans forward and falls head first without jumping. The camera tracks his accelerating descent with the sheer cliff visible behind him, then smoothly arcs as a dragon already flying through the mountain valley approaches and sweeps beneath him. He grabs the base of its neck, pulls himself onto its back and settles astride behind its shoulders. Create a natural grab and riding performance; do not inherit the previs rider’s rigid pose, automatic attachment or fixed position relative to the dragon. They are still high above the city when he grabs on. Follow @Video1’s flight path as the dragon carries him down from the mountain heights toward the rooftops, maintaining fast forward flight with ample clearance above people and traffic. Transform the previs city into an immense, breathtaking sunlit fantasy city, extending beyond the frame in every direction. Preserve the buildings’ broad spatial arrangement and the clearances required by the flight path while replacing their basic forms with intricate architecture. Dense layers of ivory stone buildings, terracotta rooftops, ornate balconies, domes, bridges and towers of varied heights surround broad avenues and narrower market streets. Warm sunlight catches the upper façades while cool shadows fill the streets below. Jagged mountains rise beyond the city through layers of atmospheric haze. The streets feel alive: pedestrians cross plazas, merchants serve customers beneath colourful awnings, carts travel along the avenues, and residents on balconies look up and wave as the dragon passes. Follow @Video1’s sideways passage between the two buildings. Establish the narrow opening in the approaching view so the bank has a clear purpose. The dragon angles its body and draws its wings inward to pass cleanly between the façades, with enough room for the rider, then levels out into a broad avenue. The rider leans naturally with the bank while maintaining his grip. Preserve the relationship between the flight path and surrounding architecture so the manoeuvre flows continuously through the gap. Nearby rooftops and balconies sweep rapidly through the foreground, while distant towers move more slowly, making the dragon’s speed tangible. Preserve @Video1’s continuous camera movement through the flight and its final sweeping reveal. As the camera widens, reveal the city’s enormous scale: countless streets, rooftops and towers stretching toward the distant mountains, with the dragon and rider remaining clearly visible as they fly away. Replace all previs assets with finished imagery: the exact character from @Image1, a fully detailed winged dragon with layered scales and natural flight animation, intricate architecture and natural mountain formations, all rendered in @Image1’s exact style. Do not inherit @Video1’s low-poly geometry, faceted surfaces, mannequin anatomy, simplified character or dragon design, placeholder materials, viewport lighting or Blender overlays. Preserve its broad choreography and camera movement while completely rebuilding its visual appearance and naturally animating the rider’s interaction with the dragon.
3
17
1,591
Grand Visions retweeted
BEYOND THE PATH (その道の先) Episode 1: The Boy Who Cut the Sky #Anime #AIAnimation #AIFilmmaking #Seedance
5
2
2
17
1,074
Grand Visions retweeted
Seedance 2.5, anime. Lots of fun (and also frustration) if you’ve got the credits. This scene looks simple, but it was ridiculously difficult to pull off. It took a lot of work and iteration (hence 480p), and still isn’t perfect, but it's starting to get expensive so I figured I'd call it here. I’m looking forward to the day I can create freely without cost forcing me to compromise on my vision. #Seedance #AIVideo #AIAnimation #AIFilmmaking
2
1
6
825
Seedance 2.5, anime. Lots of fun (and also frustration) if you’ve got the credits. This scene looks simple, but it was ridiculously difficult to pull off. It took a lot of work and iteration (hence 480p), and still isn’t perfect, but it's starting to get expensive so I figured I'd call it here. I’m looking forward to the day I can create freely without cost forcing me to compromise on my vision. #Seedance #AIVideo #AIAnimation #AIFilmmaking
2
1
6
825
Brothers reunited… Wan 3.0 @Alibaba_Wan. This John fellow wasn’t in the prompt, but he looks awfully familiar. 👀 #Wan3 #AIFilmmaking #AIVideo #Seedance
3
1
14
666
Grand Visions retweeted
We're giving away 6,000,000 credits. One mission: create. Fuel your next AI film with PixLight. Start creating. Start your story. Follow + RT + Reply = 150 Creds & Access to 6M Creds in DMs(72H ONLY)
566
525
25
578
24,251
Grand Visions retweeted
Another incredible example of what Wan 3.0 can do, this time with anime. Huge appreciation to the @Alibaba_Wan team for building something this amazing. I’d love the opportunity to work with the team in any capacity, help push the model even further, and showcase what it’s truly capable of. 🙏 And since many people have asked, I’m using Wan 3.0 through @TopviewAIhq @TopviewAIJP. My plan includes free 720p generations. It can be slow, but it’s reliable, and I consistently get at least a few videos generated each day. Not sponsored by the way. I’d love to work in 1080p and create longer form content, but that’s beyond my budget at the moment. Hopefully soon!
2
2
4
588
Another incredible example of what Wan 3.0 can do, this time with anime. Huge appreciation to the @Alibaba_Wan team for building something this amazing. I’d love the opportunity to work with the team in any capacity, help push the model even further, and showcase what it’s truly capable of. 🙏 And since many people have asked, I’m using Wan 3.0 through @TopviewAIhq @TopviewAIJP. My plan includes free 720p generations. It can be slow, but it’s reliable, and I consistently get at least a few videos generated each day. Not sponsored by the way. I’d love to work in 1080p and create longer form content, but that’s beyond my budget at the moment. Hopefully soon!
2
2
4
588
Grand Visions retweeted
My entry for Topview AI’s 30-Second Challenge. I’m not seeing many dialogue scenes in this challenge, which makes sense. The visuals are what most people care about, after all. And even in the dialogue scenes I have seen, there’s often this huge emphasis on making everything LOOK cinematic. Color grading, upscaling, extra polish, pushing the image as far as possible. All of that visual polish to keep your eyes busy enough that you don’t notice the wooden acting underneath. I guess that works for most people. It just doesn’t do much for me. Wan 3.0, you know the drill. Raw 720p. No color grading. No upscaling. I’m trying to make people feel something here, and for me, a strong performance trumps even the most visually spectacular Seedance video you could ever make. To my Japanese friends, please let me know how the Japanese sounds. It seems pretty good to me... でも、私は日本人じゃない 😅 And everyone else, let me know what you think as well! @TopviewAIhq @TopviewAIJP #Topview30sChallenge #TopviewAI
Think you can make an impact in just 30 seconds? The Topview 30s Challenge is here! 🚀 Create. Experiment. Get creative. Show us what you can do and turn 30 seconds into something unforgettable. 👉 Join the challenge: topview.ai/activity/topview-… #Topview30sChallenge
4
3
8
523
My entry for Topview AI’s 30-Second Challenge. I’m not seeing many dialogue scenes in this challenge, which makes sense. The visuals are what most people care about, after all. And even in the dialogue scenes I have seen, there’s often this huge emphasis on making everything LOOK cinematic. Color grading, upscaling, extra polish, pushing the image as far as possible. All of that visual polish to keep your eyes busy enough that you don’t notice the wooden acting underneath. I guess that works for most people. It just doesn’t do much for me. Wan 3.0, you know the drill. Raw 720p. No color grading. No upscaling. I’m trying to make people feel something here, and for me, a strong performance trumps even the most visually spectacular Seedance video you could ever make. To my Japanese friends, please let me know how the Japanese sounds. It seems pretty good to me... でも、私は日本人じゃない 😅 And everyone else, let me know what you think as well! @TopviewAIhq @TopviewAIJP #Topview30sChallenge #TopviewAI
Think you can make an impact in just 30 seconds? The Topview 30s Challenge is here! 🚀 Create. Experiment. Get creative. Show us what you can do and turn 30 seconds into something unforgettable. 👉 Join the challenge: topview.ai/activity/topview-… #Topview30sChallenge
4
3
8
523
Thank you, @sarasara_aiart さん, for your impressive Wan 3.0 submission and for inspiring me to give the challenge a go. Definitely check out their entry here: xitter.cf/sarasara_aiart/status/…
Topview 30秒チャレンジ! せっかくなのでWan3.0君のT2Vで応募です! Wan3.0君は私はいつもTopviewさんで使わせてもらってます🥰 @TopviewAIhq #Topview30sChallenge #TopviewAI
90
To be clear, I’ve got nothing against color grading or upscaling. Color grading is a skillset in its own right, and I’ll absolutely be doing both when they actually serve the shot. I just don’t think they should be used to compensate for a performance that isn’t there in the first place.
127
Wan 3.0 is, once again, truly remarkable. @henrydaubrez recently made a great point about how quickly AI filmmakers seem to settle for “good enough.” In my reply, I suggested that many people may not be consciously settling at all. They may simply not notice the problems. I wonder whether this is especially true of acting. Seedance 2.5 is regularly praised for its performances. Yet, as I’ve said many times before, although Seedance has largely solved visual realism, the moment a scene contains sustained dialogue, its wooden acting performances become evident. Wan 3.0 is levels beyond anything I’ve seen so far. It doesn’t just look real. It feels real. This is why I keep trying to spread the word about the model. I have nothing to gain from doing so, and I’m not sponsored by Wan or anyone else. Of course, this video is not perfect. It contains continuity errors and other flaws. But if I’m immersed in a story, I barely notice minor continuity issues. A wooden acting performance, however, pulls me out immediately. How sensitive are you to acting in AI video? Do you notice wooden dialogue performances but accept them as “good enough,” or do they not break the illusion for you? If you’ve seen an AI acting performance that genuinely feels human, please share it below. #Wan3 #Seedance #AIFilmmaking #AIVideo
3
1
10
403
Henry's post, worth a read. xitter.cf/henrydaubrez/status/20…
One of my biggest pet peeves with AI filmmaking right now is how quickly we seem to settle for “good enough.” We keep saying these tools remove friction without lowering quality, and yet I still see so many 15-second sequences glued together with broken continuity, disappearing objects, obvious artifacts, no grading, barely any dialogue, and very little attempt at actual narration. The funny part is that the tools are getting insanely good, but somehow that seems to make some of us less demanding, not more. Getting a cool shot fast shouldn’t be the end goal. It should give us more time to curate, redo the bad stuff, fix continuity, work on performance, edit properly, and actually tell a story. Nothing wrong with short experiments or gimmicks, I make them too. But if we want people to take AI filmmaking seriously, we also have to show that we care about the filmmaking part, not just the generation. And then we wonder why the whole space gets dismissed so easily. None of us will get it perfect, myself included, but we can at least give enough of a shit to keep raising the bar.
123
Henry's post, worth a read. xitter.cf/henrydaubrez/status/20…
One of my biggest pet peeves with AI filmmaking right now is how quickly we seem to settle for “good enough.” We keep saying these tools remove friction without lowering quality, and yet I still see so many 15-second sequences glued together with broken continuity, disappearing objects, obvious artifacts, no grading, barely any dialogue, and very little attempt at actual narration. The funny part is that the tools are getting insanely good, but somehow that seems to make some of us less demanding, not more. Getting a cool shot fast shouldn’t be the end goal. It should give us more time to curate, redo the bad stuff, fix continuity, work on performance, edit properly, and actually tell a story. Nothing wrong with short experiments or gimmicks, I make them too. But if we want people to take AI filmmaking seriously, we also have to show that we care about the filmmaking part, not just the generation. And then we wonder why the whole space gets dismissed so easily. None of us will get it perfect, myself included, but we can at least give enough of a shit to keep raising the bar.
7