I’ve watched a bunch of these and this is the best showcase of minimax so far imo. Very clean and well done. This short scene is engaging and makes me want to watch the rest of it if it was a real movie/show!
Mini Max is currently taking off and in my opinion, it can be said that it is the first artificial intelligence of filmmaking through text writing, which is really powerful, very intelligent and especially creative, it is moving forward, especially on cinematic techniques, it is very wonderful.
@@onlineispections I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
The acting of the main character is superb. His gait is perfect. His hesitation to go to the house is palpable. Great establishing shots. Beautiful atmosphere. Great soundtrack. I really wanted to know what happened next!
I make AI short films and was impressed by your film. MiniMax is great but such a shame it doesn’t have image-to-video yet. Good job with the consistency of the character in your film! I bet it took a lot of rerolls to achieve the final shots and the film!
@@TeemuKetola Thanks so much :) I didn't have to do many iterations, in fact, I think I was mostly lucky 😅 but sometimes my character was bald and beardless. The plans were pretty but unusable because they were incoherent haha.
@@florianlepriol Nice to hear that you had luck with it! :D I'm currently trying to create some b-roll with MiniMax for my next AI film and I'm not that lucky, having to create about 10 to 15 shots to get one decent one haha
I was thinking the exact same thing that consistency is there even though they have no consistency controls you have to promise really well to make sure they look the same on each shot. kudos!
@@florianlepriol I love it too!! But could you achieve this character and style consistency using directly text-to-image? It's truly amazing man! Could you share some of the prompts or the structure of them? Just yesterday I uploaded my first video on Minimax in which I use the image-to-video capability to create highly realistic videos -> ua-cam.com/video/sOQ5dytqyBM/v-deo.html Quite happy with the results there but I'm still learning how to maximize quality with text-to-video. I would appreciate some tips :)
@@jesuscopado-en Thank you so much for your kind words! 😊 Achieving character and style consistency directly with text-to-image can be challenging but absolutely doable with structured prompts. Here's an example of one of my prompts to give you an idea: "Generate a six-second high-resolution video showing a close-up of the hand of a stocky, 40-year-old bearded detective as he knocks vigorously on the weathered wooden door of a chalet. He is wearing a dark coat and a worn-out hat, both adding to his rugged appearance. The door's old, peeling paint and rough texture are highlighted in the shot, emphasizing its age and exposure to the elements. The camera captures the forceful impact of his hand, with the atmosphere around being gray, misty, and damp but not actively raining. The setting is somber and foreboding, echoing the overall mood of isolation and suspense." If you look closely at my video, you'll notice that the character isn't perfectly consistent-especially when it comes to the hats! This is a common challenge when working with AI-generated outputs. To improve consistency, the detailed description of the character you use in each prompt is very important. You need to find a description that works well for your character and run tests to refine it. Once the AI has a good grasp of your character, you can start adding details about the environment, camera, etc. One great tip is to have a chat with ChatGPT, teach it what your character looks like, and even give your character a name. That way, you can ask ChatGPT to help you craft prompts with the same detailed character description each time. This helps maintain consistency across your creations. Keep testing, and don't hesitate to reach out if you need more tips or want feedback on your results!
@@florianlepriol Wow, this is golden, appreciate this a lot :) I'm already using ChatGPT for my Minimax prompts but didn't think of teaching it about the character with a name, that's a neat idea!! My next challenge is to build a cinematic short like yours with the same character, but I actually was thinking of training first a LoRA (FLUX) to learn about my character, so that he/she could be somehow less characteristic and then use this LoRA for all base images. What do you think of this approach? Couple more questions if you don't mind :D. Did you upscale the videos with Topaz or similar or these are directly as they come from Minimax? And did you color grade them? I love the overall aesthetic you achieved, I completely missed the detail about the slightly different hats the character uses cause how good the style consistency is across the scenes!
@@jesuscopado-en Thanks a lot, I’m glad you found it helpful 😊 For this video, the size and quality are directly from Minimax. However, I recently tested TOPAZ for upscaling AI videos, and it’s pretty magical-it really enhances the resolution nicely. There’s also VideoProc Converter AI, which works quite well for upscaling too. For color grading, I use Adobe Premiere with the Lumetri tool to bring out specific tones and create that cinematic feel. This really helps unify the aesthetic across scenes. Creating a LoRA is definitely a great way to keep your character coherent. I’m planning to explore FLUX and LoRA creation myself, as I believe this is the way forward. The goal is to create a LoRA for my characters, generate cinematic frames with FLUX assisted by ChatGPT prompts, and then animate everything with Minimax. Thanks for noticing the style consistency-it’s something I focus on, even with slight differences like the hats! Let me know if you have more questions, happy to help!
The AI model is superb, especially it's ability to use images as a reference for the entire three dimensional scene/objects in them, but sadly the bitrate of the downloaded videos is Tik-Tok/UA-cam quality at best for scenes with a lot going on so you get loads of blocking up on the textures using my complex highly realistic scene prompts. Great start though and the best 'free trial' of the big 4 AI video options worth using.
Thanks ! I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
Thanks ! I write a brief description of the character and the mood that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
In the foreground, a stocky detective stands, his broad frame and heavy trench coat making him appear even larger. He leans slightly, favoring one leg, showing discomfort from an old injury that still bothers him. His weathered face shows years of hard work, stress, and experience. The camera focuses on his intense expression, as his deep-set eyes scan the environment with caution. His hand hovers near the pocket of his coat, where a notepad or weapon might be hidden. The atmosphere is dark and moody, with dim lighting reflecting off the rain-soaked streets in the background. A faint mist rises from the ground, adding to the noir, gritty feel of the scene.
Thanks ! I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
@@gwart41 Thanks ! I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3. One exemple for a prompt : a 50-year-old bearded detective, with a dark jacket and a hat, slightly fat body.
@@florianlepriol thanks alot, your consistency with the character really is great! For mine I used Leonardo dicaprio so I didnt have to describe him lol. ua-cam.com/video/fWk8JC_pXjg/v-deo.html
@@Insertrandomnamehere412 on Adobe Premiere pro and with sounds found on the internet, I'll let you go see the site called "Freesound" which is perfect, you just need to subscribe and you can have all the sounds you want in super quality :)
Thanks ! I write a brief description of the character and the mood that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
I’ve watched a bunch of these and this is the best showcase of minimax so far imo. Very clean and well done. This short scene is engaging and makes me want to watch the rest of it if it was a real movie/show!
@@HJT1 Thanks so much ! Your words motivates me a lot to continue making other tests and micro films !
Mini Max is currently taking off and in my opinion, it can be said that it is the first artificial intelligence of filmmaking through text writing, which is really powerful, very intelligent and especially creative, it is moving forward, especially on cinematic techniques, it is very wonderful.
Hi. How did you make consubstantia character before there was no image to video.
@@onlineispections I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
The number of minimax users has recently increased quite a lot. Perhaps that is also a signal showing the power of this AI in making videos.
The acting of the main character is superb. His gait is perfect. His hesitation to go to the house is palpable. Great establishing shots. Beautiful atmosphere. Great soundtrack. I really wanted to know what happened next!
Thanks so much !
in 3-5 years AI will go beyond than holywood shlt bouto get real its just matter of time now
I make AI short films and was impressed by your film. MiniMax is great but such a shame it doesn’t have image-to-video yet. Good job with the consistency of the character in your film! I bet it took a lot of rerolls to achieve the final shots and the film!
@@TeemuKetola Thanks so much :) I didn't have to do many iterations, in fact, I think I was mostly lucky 😅 but sometimes my character was bald and beardless. The plans were pretty but unusable because they were incoherent haha.
@@florianlepriol Nice to hear that you had luck with it! :D I'm currently trying to create some b-roll with MiniMax for my next AI film and I'm not that lucky, having to create about 10 to 15 shots to get one decent one haha
Image to video according to their discord is coming out in less than a week
@@gwart41That’s cool! Can’t wait :)
I was thinking the exact same thing that consistency is there even though they have no consistency controls you have to promise really well to make sure they look the same on each shot. kudos!
This is fantastic Florian, amazing work!
Thanks so much ! I really love Minimax and its ability to understand prompts and render everything in a fluid, cinematic way.
@@florianlepriol I love it too!! But could you achieve this character and style consistency using directly text-to-image? It's truly amazing man! Could you share some of the prompts or the structure of them? Just yesterday I uploaded my first video on Minimax in which I use the image-to-video capability to create highly realistic videos -> ua-cam.com/video/sOQ5dytqyBM/v-deo.html
Quite happy with the results there but I'm still learning how to maximize quality with text-to-video. I would appreciate some tips :)
@@jesuscopado-en Thank you so much for your kind words! 😊 Achieving character and style consistency directly with text-to-image can be challenging but absolutely doable with structured prompts. Here's an example of one of my prompts to give you an idea:
"Generate a six-second high-resolution video showing a close-up of the hand of a stocky, 40-year-old bearded detective as he knocks vigorously on the weathered wooden door of a chalet. He is wearing a dark coat and a worn-out hat, both adding to his rugged appearance. The door's old, peeling paint and rough texture are highlighted in the shot, emphasizing its age and exposure to the elements. The camera captures the forceful impact of his hand, with the atmosphere around being gray, misty, and damp but not actively raining. The setting is somber and foreboding, echoing the overall mood of isolation and suspense."
If you look closely at my video, you'll notice that the character isn't perfectly consistent-especially when it comes to the hats! This is a common challenge when working with AI-generated outputs.
To improve consistency, the detailed description of the character you use in each prompt is very important. You need to find a description that works well for your character and run tests to refine it. Once the AI has a good grasp of your character, you can start adding details about the environment, camera, etc.
One great tip is to have a chat with ChatGPT, teach it what your character looks like, and even give your character a name. That way, you can ask ChatGPT to help you craft prompts with the same detailed character description each time. This helps maintain consistency across your creations.
Keep testing, and don't hesitate to reach out if you need more tips or want feedback on your results!
@@florianlepriol Wow, this is golden, appreciate this a lot :)
I'm already using ChatGPT for my Minimax prompts but didn't think of teaching it about the character with a name, that's a neat idea!!
My next challenge is to build a cinematic short like yours with the same character, but I actually was thinking of training first a LoRA (FLUX) to learn about my character, so that he/she could be somehow less characteristic and then use this LoRA for all base images. What do you think of this approach?
Couple more questions if you don't mind :D. Did you upscale the videos with Topaz or similar or these are directly as they come from Minimax? And did you color grade them? I love the overall aesthetic you achieved, I completely missed the detail about the slightly different hats the character uses cause how good the style consistency is across the scenes!
@@jesuscopado-en Thanks a lot, I’m glad you found it helpful 😊 For this video, the size and quality are directly from Minimax. However, I recently tested TOPAZ for upscaling AI videos, and it’s pretty magical-it really enhances the resolution nicely. There’s also VideoProc Converter AI, which works quite well for upscaling too.
For color grading, I use Adobe Premiere with the Lumetri tool to bring out specific tones and create that cinematic feel. This really helps unify the aesthetic across scenes.
Creating a LoRA is definitely a great way to keep your character coherent. I’m planning to explore FLUX and LoRA creation myself, as I believe this is the way forward. The goal is to create a LoRA for my characters, generate cinematic frames with FLUX assisted by ChatGPT prompts, and then animate everything with Minimax.
Thanks for noticing the style consistency-it’s something I focus on, even with slight differences like the hats! Let me know if you have more questions, happy to help!
Superb, Florian, congratulations. The character consistency is excellent, as is the lighting style.
Thanks !
The AI model is superb, especially it's ability to use images as a reference for the entire three dimensional scene/objects in them, but sadly the bitrate of the downloaded videos is Tik-Tok/UA-cam quality at best for scenes with a lot going on so you get loads of blocking up on the textures using my complex highly realistic scene prompts. Great start though and the best 'free trial' of the big 4 AI video options worth using.
How did you keep the character consistent? It looks great
Thanks ! I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
Thanks for the info. Looking forward to seeing what else you create. Great work.
but how u use the sound effect? like knocking door, environtment, etc. Is that manual/ai?
For the sound design, I used sounds recorded with my own microphone, my personal sound bank and the “Freesound” website :)
If I may ask, how did you keep a fairly consistent character in MiniMax with out image to video. This is well done!
Thanks ! I write a brief description of the character and the mood that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
Can you share the sample of your prompt ? Which you used for consistent footage
In the foreground, a stocky detective stands, his broad frame and heavy trench coat making him appear even larger. He leans slightly, favoring one leg, showing discomfort from an old injury that still bothers him. His weathered face shows years of hard work, stress, and experience. The camera focuses on his intense expression, as his deep-set eyes scan the environment with caution. His hand hovers near the pocket of his coat, where a notepad or weapon might be hidden. The atmosphere is dark and moody, with dim lighting reflecting off the rain-soaked streets in the background. A faint mist rises from the ground, adding to the noir, gritty feel of the scene.
but I've just had a lot of luck, I think, with the consistency of the character !
Great work! How on earth did you get character consistency!? incredible!
Thanks ! I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
I want api for minimax and character consistency. also, complete movie machine that also adds cloned voiced from script. one prompt to complete movie.
Are you sure this is not filmed? Very impressive
@@Sergio-Sanchez-comyes i'm sure haha , thanks !
Amazing, Amazing Work!
Thanks !
Watching this video is no different from being filmed by the filmmaker.❤
Thanks !
How you make consistent character?
What was the prompt you used to get the consistent man in the hat and leather jacket?
@@gwart41 Thanks ! I write a brief description of the character that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3. One exemple for a prompt : a 50-year-old bearded detective, with a dark jacket and a hat, slightly fat body.
@@florianlepriol thanks alot, your consistency with the character really is great! For mine I used Leonardo dicaprio so I didnt have to describe him lol. ua-cam.com/video/fWk8JC_pXjg/v-deo.html
Where did you make the sound effects?
Elevenlabs
@@Insertrandomnamehere412 on Adobe Premiere pro and with sounds found on the internet, I'll let you go see the site called "Freesound" which is perfect, you just need to subscribe and you can have all the sounds you want in super quality :)
unbelievable, but how do you explain, same light, same character etc.?
Thanks ! I write a brief description of the character and the mood that I put in each prompt, but it doesn't always work, for example sometimes minimax would put me a man without a beard and a thin body. But I think with a good physical description there's a way to keep good consistency like in gen-3
@@florianlepriol thanks
Where will humans gonna go? Then
You have a good taste 😊 the sound effects are not elevenlabs right? Timing is perfect as well 👍
Thanks ! No elevenlabs here, personal sound bank + the website " Freesound " :)
@@florianlepriol perfect 👍 Love that!
complimenti
Thanks !
Where is watermark?
Good job
Thanks !
I signed up immediately
Only 6 Seconds videos 🤦🏽♂️ too bad
I have created my first video with this AI tool. Please check and comment