🔥 A hot tip: 🔥 If you want to animate your own image, then animate it and take a screenshot of the last frame of the video. Scale up the screenshot and animate the image again. This gives you a clean flow and allows you to produce longer videos if you want to create a film.
@@IIIIIOIIIIII Since it will then be a new scene, you will of course have to enter a new prompt. However, I usually only change the part of the prompt that the character does. Of course you can leave all other settings such as “light”, “atmosphere”.
Thx a lot for the tip and Tim for the video! This might be the best way to keep at least some consistency for a longer video built with lot short clips!
@@TheoreticallyMedia Yes but I tried generating 3 videos in Kling, one of them successful, 2 processed for hours and after that gave a "Process failed"
@@shredd5705I've gotten plenty of can't process, but I've also gotten really good results. Playing with minimax now. Looks like it's not maintaining the image well.
haha THANKS Tim for the shoutout - the crazy thing about that generation is that Minimax was able to (correctly) predict what was behind my daughter's painting (a black oven) - which was not even visible in the original input image! ::mindblown::
That is CRAZY. I also think it’s super cheeky of Minimax to have your daughter flip the painting as opposed to animating what is already there. Almost like MM was like, “that’s too much work, I know what I can do!” Great generation though, and such and excellent and creative idea! Your daughter’s painting is great as well!
@@TheoreticallyMediaagree. As an aside I actuallly was hoping that mm would animate the painting she was holding, and even tried several times to make additional generations (with prompt this time trying to animate the painting)… but it wouldn’t do it (only tried about 6 times)
I think the next level for these generators will be to create their scenes in fully formed 3d assets to maintain 100% consistency and physics fidelity. The current 2d techniques would be used post "production" to polish the 3d modeled outcome to a photoreal finished product.
Yeah, I agree. In the meantime, I was wondering about creating 3D stuff in Daz Studio and applying AI gen video techniques to it to polish it off for a photoreal finished product, since that's what I need for my purposes. Possible?
I don't mean to pick on Gen-3 either but MiniMax blows it away in many ways. There's just a lot Gen-3 can't do, where Kling and even Luma still can. In fact I'm actually a little disappointed in Runways long in the works Gen3, especially using their own generator which is pretty terrible much of the time. Luma needs an upgrade though, I swear they were producing slightly better results when they first launched at least in terms of speed, they've taken a small step back with their last update. Pika's update is,..interesting but not up to par, not even close. MiniMax and Kling are by far the leaders, thank God for the Chinese though Kling is INSANELY expensive and I fear MiniMax will be as well once they go to pay. They will have the last laugh when they drain our wallets
@@phu320What the hell is this propaganda going on? Who says that? Imagine saying “thank god for the Americans” when complimenting Runway or Luma 😂 what does race have to do anything with this? Are you guys communist bots??
It's funny, one of the reasons I do the "guy in a blue business suit" test in most of these videos is because it's kind of a "bland" example. But, I think it's a great showcase of what these models can do as a baseline. Stuff like "Super Sci-fi Adventure with Zombies and Hot Women With Laser Katanas" is great and all-- but, if the model can't handle a guy walking down a street, it probably isn't going to do that Sci-Fi epic very well!
I feel it is not just the best free one. When it comes to coherency and following prompts even with image 2 video it is a big step ahead. I for example used an image of my cat to create a 5 second clip and then I used the last frame and did it over and over again and yet the cat did not morph to something else like with runway or luma.
I think I should have time to pull it off next week! It’s funny, isn’t AI supposed to make things “easy”and you just prompt it into ChatGPT? Sigh. Sarcasm obviously!
Yeah, I've been getting a lot of comments where people want me to explore outside of the usual cinematic type stuff. Always happy to incorporate when I can!
Definitely been waiting for this upgrade. Kinda have a thing for Minimax. Will put it to the grind first thing in the morning. Hopefully, we will come up with something cool. Great video, Tim.
AI video has gone from a dangerous crawl to exponential growth in a year. 8:10 Yeah, the changing of the focal point was very impressive. Thanks again, Tim. 💯
Great video! I just switched to Minimax from Luna Dream Machine and it is night and day! Prices are better and videos are 10x better. Only down side is the speed of Minimax.
I just tried it before watching this video and was pleasantly surprised they finally added this feature and omg, the results are superior to anything I've tried before, hope it stays that way, even if it doesn't get better!!
Whatr I have found is Minimax is extremley good at logo graphics, very good at spelling and has produced some good concepts for me without using an image just a good prompt
@@TheoreticallyMedia Yeah thanks, makes sense. I did 24 videos yesterday, out of the 24, maybe 12 seconds worth of decent footage (being picky). It will be interesting to see how that changes with image-to-vid. On average, I'd say between 5-7 minutes per generation at various times of the day. I like that it tells you, how many are ahead of you.
I've had those same issues with Gen 3 img2vid where it completely loses the style. It seems to work better for landscapes. After playing around with minimax img2vid I think it's better than Runway's, but Kling 1.5 is still king. It keeps surprising me with how well it understands the prompt and maintains the same style.
This is an interesting development. Tim, sorry, little off-topic, but I wonder - have you ever done a video on those historical portraits to life like animated video channels? I am going crazy trying to figure out which tools they are using to do that stuff. Would love to experiment with it myself.
I purchased a month of unlimited Runway, I'm looking forward to more control over characters because when it gets to that detailed you will be able to create a whole movie from image prompts
Sub just lightly announced! I've got a community post up on it! looks like we'll still have free, but also a standard at 9 and a unlimited at 94...big jump there!
I welcome you. The video is very good, it helped a lot to navigate. I would like to ask how it can be set in the text description so that the video does not move in several directions? I mean that it should be in a side view and in a fixed position.Thank you.
Hi thanks for the video. With luma you can make sequence shots, add video. In minimax the last frame. When will it be possible to do a three-minute scene where two actors are talking in a room, with different camera angles? Shot, reverse shot, title, etc.?
@@TheoreticallyMedia Hi I wanted to know from you if luma works because here it doesn't work anymore at one point I wanted to ask you if to make Mira Max work without subscriptions you can create new Google accounts But do you know that they also ask for the phone number to create new Google accounts to make Mira Max work for free? Hi
I love Morph-Fu. And your Jersey cap. As soon as MiniMax went image to vid, I was in the frickin queue. I took a screenie and then prompted a new action and it works so well. I'm so excited!
I KNOW! I love Morph-Fu too! I really think it could be used to create a cool sequence, if you built a story around it. Like a fight scene in a dream! Hmmmm....Might need to ponder that!
Tim, your GPT for MiniMax is the best thing since sliced bread. By far it's the best prompt generator I've worked with. Even the English output gives me near perfect results every clip! It's a shame you can't monetize it. I'm about 140 clips into a large project and almost all of it is usable in some shape or form. Too bad we can't get rid of the watermark though.
YOU slayed me Tim. Driver from Joisy. OMG. Yep Big Apple knows. Part time at stand up Tim? I'd go see ya. Great stuff as always. I'd pay for Minimax, look forward to it. They consistently produce the best text to image and I've tried them all.
Should be mentioned that with the free plan you get a watermark on your videos. And compared with Runway Gen-3 I get worse result, but these are the very first tests, maybe minimax needs different kind of prompting than runway.
Well this morning they announced on discord: "Tomorrow, we are launching new paid plans designed to elevate your creative experience! Enjoy perks like unlimited credits, faster generation times, queuing multiple tasks, watermark removal, and priority access to the latest feature". It was a really FUN ride.... we will see what happens tomorrow LOL At this point SORA and Movie Gen are you going to have to be 10x better and free for weeks to catch up to the Kling and Minimax sales market. So that might be a WIN for us lol. THANKS again Tim for the Prompt!! Awesome job! Really glad I found you and subscribed at the start of the Pika era. Keep up the good work!
Tim so out of all of them, Klingon, runway, pika , minimax etc, which one do you think overall is the best? I would love to see a shootout video reciew
So, to be honest, I try to avoid doing those "shootout" and "this killed this" kind of videos. I know those things tend to get big numbers on YT, but I also think they can skew perspective depending on what is being tested. I feel like all of them excel at certain things-- and if there's one thing I've learned from doing these videos, it's that any one of them are capable of flipping the table with one update. All that said, you did get me thinking about maybe doing a video showcasing what each one of them is great at. THAT might be an interesting video-- and kind of a positive spin on the whole thing!
Great video! Any tips on how to make the characters seem to be silent and not talking? Cause most of the time, they start yapping ceaselessly for me haha!
01:48 The 6-second number may be rooted in Chinese culture where 6-8-9 are good luck numbers (4 is the worst and always avoided). Funny thing this is a Chinese tool but this new feature is not yet available here in China. Hopefully is in the next hours, I can't wait!
Ohhh, that IS interesting!! and-- makes sense! I wonder if in the states, we'll never get a 13 second generation! Also, I didn't realize Minimax hasn't released I2V in China-- BUT, you guys DO get Bytedance's Seaweed-- so, you've got a bit of a win there!
Thankyou! Thankyou Thankyou! Tried this out and used the prompts. This is gonna be so fun! Thankyou for your knowledge and time you took to make this video.
The 3d stuff was pretty good, but yeah, that one cyberpunkish girl, I did note that she sort of remained 2Dish? I can see that being something the model isn’t great with
Try to use this at the end of each prompt. It works for me: No ARTIFACTS, SLOW, UGLY, BLURRY, DEFORMED, MULTIPLE LIMBS, CARTOON, ANIME, PIXELATED, STATIC, FOG, FLAT, UNCLEAR, DISTORTED, ERROR, STILL, LOW RESOLUTION, OVERSATURATED, GRAIN, BLUR, MORPHING, WARPING.
Totally! and it's getting close-- The characters aren't teleporting all over the place. I'd say give it a few more months, and we might see something moving that dynamically, without the decoherence!
Get weird Tim! Experimental stuff is where the real good work comes from. Super helpful and informative as always. This is the ai visual media channel I send anyone to who is doing real work. With all of these advances though, I can't help but wonder...and be amused...that by the time this stuff is 'perfected' ... nobody is going to care about movies anymore. Between quality of direction, attention span, writing, hollywoodism being so prevalent, and fatigue, its possible we end up where people just want ... physical art and jewelry lol. We'll see what we end up with as corps bleed us dry.
O yea btw 7:00 min. In you say that there's a large percent of us that dislikes when you get all experimental, well i for one am not in that # my friend. I in fact enjoy when you get experimental. 😊
Thank you so much for keeping on point with all the ai-ness. The thing for me with Minimax and the free tier is that you get 100 free tokens a day and that allows you to create 3 videos at 30 tokens a shot with 10 tokens left over that don't roll into the next day 😂. But nonetheless, onward and upward!!!!
Thank you! If you're interested in seeing what advanced AI video generators can do, I’ve made a fully AI-generated 3-minute music video on my channel, along with a 1-minute short created using LTX Studio, for comparison. I’m not sure if I’ll be able to achieve the same results with Minimax, but I’m definitely going to give it a try.
I heard that as of tomorrow MiniMax will cost $95 per month for unlimited generations. Same as Runway Gen 3 Alpha, which does so much more. Is this still a good deal? Also, Adobe's Firefly Video is dropping sometime this year, and it's included with the Creative Cloud subscription.
Nice! I've been pretty happy with Runway ML, but Hailuo is ...maybe better? So far I've only used it to animate an AI-generated home studio that I plan to use with a green screen. I wanted to animate the images on the computer monitors, and other very minor movement of foliage WITHOUT camera movement. Minimax seems better than Runway at this. So far, characters aren't working as well for me as Runway. When my character's head turns, it warps, and doesn't exactly keep the same features. Also, the teeth look strange. But still impressive, and I love that I have got a loopable background for a green screen set.
can you plz explain how to continue to further shots with same scene, consistent character, and camera moves when we are creating a movie? because all the view gets changed for next shot, and sometimes character as well. Every expert on youtube is showing just 1 shot. Please let us learn to shoot a scene of 6 shots, 3 minutes long, showing 7 friends. Refer to any sample scene from series FRIENDS. Many thanks
current fighting generated videos remind me most of some dragon ball z fights when they are moving so fast it looks like multi armed blur , except unfortunately that warbling floaty slomo action of generated videos feels different , if you speed it up tho and drop frames like animating on 2s it can be neat
Great video! Thanks for the heads up! I think the future of AI will be dozens of different #AIVideo sites, each with their own style specialization. We need a series of "real life" LLMs and then one "anime", one "manga", one "cartoon", and so forth.
@@harrygoldhagen2732 So, I agree. As someone with three comic book scripts burning a hole in a pocket, that would be lovely. However, here's the tricky part. When you take a photograph of a human, there is no way to prevent people from duplicating your work if they had the same model, setting, so forth. But when you draw, your art style is uniquely yours. And the second you "give" the style to AI, you're out of work. It would be wonderful to have such AI, but you would be asking for a great many sacrifices and I'm not sure that's feasible.
@@dirtydevotee Very true! But the frustration of having many scripts -- I have a bunch! -- just sitting there undrawn, makes me want to have some tool to make them faster.
@@harrygoldhagen2732 I would like to offer that there are LoRAs you can pair with either Dezgo or Rubbrband that will get you a consistent art style. 90% of them are free. Inzaniak has a fantastic "Graphic Novel - Style" LoRA for SD 1.5 on Civit. Perhaps that might help you edge a little closer.
Viewers are shown examples of how the image-to-video feature works, noting that it can sometimes skew towards either a cartoonish or realistic style, depending on the input image. Thank you
Are you only meant to use parts of the community prompt? As if you use it all then you basically have no space to type what you actually want in the video.
you mean from the GPT? Yeah, someimes chatGPT gets stupid and goes over the max character amount. In that case, I'll sometimes do a little editing. But, to be honest: I've run it "as is" and gotten some pretty decent results as well. It is a bit of a keyword shotgun spray, and with 2000 some odd characters, there is usually enough.
D-AI-niel san, I will teach you the secret arts of Morph-Fu :) Jokes aside, I both enjoy seeing the "Morph-Fu" sequences (they are funny) and think it's a REALLY good benchmark for fast and accurate character movements, so keep them coming! Thanks as always for your videos
the differences to runway are sensational, especially in the interpolation of the movements, in the pixa-like girl video and the godzilla video. Normally, the movement in all video generators looked far too slow and not natural
Minimax is greatm especially for free and while paid options will appear, I hope they can keep the free part going as it is. Although, I'd pay for longer vids with consistency. It's kinda funny hearing you mention that OpenAI strike as that happens to me with Instagram. I sometimes get strikes for 'spam', even though I've never made a single comment. Not one, as I only use it for movie /TV updates and info. I do the appeal thing and I'm back within minutes, so more of an annoyance than a problem. Makes me think it's a bot screwing up.
Ahhh, sorry to hear, but glad I'm not alone! And yeah, 100% suspect some sort of automation issue. I also think on the ChatGPT side, it might get tricked into giving a few keywords on the no-no list-- and considering that GPT has over 10k uses, I think they likely add up after awhile to an automated strike. It isn't a huge deal, but it always seems to happen right when I need it!
I NEED HELP IN USING AI TO GENERATE A 6 EPISODE LIMITED SERIES (EACH EPISODE A HALF HOUR OR BIT LONGER). Scripts are completed. QUESTION. CAN I UPLOAD A COMPLETED SCRIPT TO MINIMAX OR ANY OTHER AI GENERATOR...I HAVE FULLY DEVELOPED CHARACTERS, DIALOGUE, RECOGNIZABLE FOREIGN COUNTRIES LOCATIONS. AM I BEING PREMATURE? IS AI THATY ADVANCED?
Runwayml: cinematic trailers Kling AI: Cinematic and Action. Also gaming style graphics. Minimax: Action scenes with some distortion. We are almost getting there.
🔥Thank you! 🔥 to Theoretically Media for the fantastic content in your video, The Best FREE AI Video Generator Just Got BETTER! Your insights and thorough explanations made it so much easier to understand the updates and capabilities of this tool. I truly appreciate the effort you put into creating such valuable content. It’s incredibly helpful for those of us looking to enhance our video projects with AI technology!🔥Thank you! 🔥
yeah, this is going to help a lot with character consistency. So far I am screenshotting the subject of the prior video, then cropping, uploading and reprompting to continue the scene.
yes, so far so well, but still no nsfw generations possible! Worthless for me. All AI generators need to address this issue in due time. Thanks for the update, though.
🔥 A hot tip: 🔥
If you want to animate your own image, then animate it and take a screenshot of the last frame of the video. Scale up the screenshot and animate the image again. This gives you a clean flow and allows you to produce longer videos if you want to create a film.
What about the prompt? Make a new one, use the original one, or leave it blank?
@@IIIIIOIIIIII Since it will then be a new scene, you will of course have to enter a new prompt. However, I usually only change the part of the prompt that the character does. Of course you can leave all other settings such as “light”, “atmosphere”.
Cool tip 💯
The thing about AI is that it lacks consistency from point A to B. Its strenght is precisely its weakness. Too many variables.
Thx a lot for the tip and Tim for the video! This might be the best way to keep at least some consistency for a longer video built with lot short clips!
Kling doesn't work, Pika doesn't work. This works.. the results are scary good
I like Kling, but point taken: and free!
@@TheoreticallyMedia Yes but I tried generating 3 videos in Kling, one of them successful, 2 processed for hours and after that gave a "Process failed"
kling has picture to motion tho,its hell of expensive tho
@@shredd5705I've gotten plenty of can't process, but I've also gotten really good results. Playing with minimax now. Looks like it's not maintaining the image well.
@@shredd5705 not for long thanks
haha THANKS Tim for the shoutout - the crazy thing about that generation is that Minimax was able to (correctly) predict what was behind my daughter's painting (a black oven) - which was not even visible in the original input image! ::mindblown::
It knows 😳
That is CRAZY. I also think it’s super cheeky of Minimax to have your daughter flip the painting as opposed to animating what is already there. Almost like MM was like, “that’s too much work, I know what I can do!”
Great generation though, and such and excellent and creative idea!
Your daughter’s painting is great as well!
@@TheoreticallyMediaagree. As an aside I actuallly was hoping that mm would animate the painting she was holding, and even tried several times to make additional generations (with prompt this time trying to animate the painting)… but it wouldn’t do it (only tried about 6 times)
Amazing how quickly we are approaching the minor nitpick stage of AI video generator reviews.
long way from will smith eating lol
would you love to get an AI video generator?
We don't have action. It's pretty bad at it
I think the next level for these generators will be to create their scenes in fully formed 3d assets to maintain 100% consistency and physics fidelity. The current 2d techniques would be used post "production" to polish the 3d modeled outcome to a photoreal finished product.
Yeah, I agree. In the meantime, I was wondering about creating 3D stuff in Daz Studio and applying AI gen video techniques to it to polish it off for a photoreal finished product, since that's what I need for my purposes. Possible?
you are so right about Minimax being the most impressive and low-cost (so far) video generator, right now!
Its not only the best free option now. I am impressed how minimax understands and implements the images without text
I don't mean to pick on Gen-3 either but MiniMax blows it away in many ways. There's just a lot Gen-3 can't do, where Kling and even Luma still can. In fact I'm actually a little disappointed in Runways long in the works Gen3, especially using their own generator which is pretty terrible much of the time. Luma needs an upgrade though, I swear they were producing slightly better results when they first launched at least in terms of speed, they've taken a small step back with their last update. Pika's update is,..interesting but not up to par, not even close. MiniMax and Kling are by far the leaders, thank God for the Chinese though Kling is INSANELY expensive and I fear MiniMax will be as well once they go to pay. They will have the last laugh when they drain our wallets
thank God for the Chinese
@@phu320What the hell is this propaganda going on? Who says that? Imagine saying “thank god for the Americans” when complimenting Runway or Luma 😂 what does race have to do anything with this?
Are you guys communist bots??
🎯
ok the suit guy walking was seriously impressive, I'm so excited to put this to the test. Finally we have image to video in a semi-decent model.
It's funny, one of the reasons I do the "guy in a blue business suit" test in most of these videos is because it's kind of a "bland" example. But, I think it's a great showcase of what these models can do as a baseline.
Stuff like "Super Sci-fi Adventure with Zombies and Hot Women With Laser Katanas" is great and all-- but, if the model can't handle a guy walking down a street, it probably isn't going to do that Sci-Fi epic very well!
I have found one better and its crazy nobody is talking bout it
@@Avalone-w6l name?
Your tips on maintaining character consistency throughout a video we're super helpful! Thanks.
Oh, that’s awesome Dylan!! Thanks so much! We gotta catch up soon!
I feel it is not just the best free one. When it comes to coherency and following prompts even with image 2 video it is a big step ahead.
I for example used an image of my cat to create a 5 second clip and then I used the last frame and did it over and over again and yet the cat did not morph to something else like with runway or luma.
As a fan, that shot of Bruce Lee had me anticipating another film. Incredible. Good luck to everyone. Thanks, Tim.
I think I should have time to pull it off next week! It’s funny, isn’t AI supposed to make things “easy”and you just prompt it into ChatGPT? Sigh. Sarcasm obviously!
I appreciate that you also demonstrated with stylized characters. That’s what I wanted to see.
Yeah, I've been getting a lot of comments where people want me to explore outside of the usual cinematic type stuff. Always happy to incorporate when I can!
Definitely been waiting for this upgrade. Kinda have a thing for Minimax. Will put it to the grind first thing in the morning. Hopefully, we will come up with something cool. Great video, Tim.
Best bet: FIRST thing in the morning. For sure noticed a slow down around 10am as more people came online. 7am was pretty quick!
We are in Prague. So, hopefully, everyone in the US will be sleeping.
would you love to get an AI video generator?
AI video has gone from a dangerous crawl to exponential growth in a year. 8:10 Yeah, the changing of the focal point was very impressive. Thanks again, Tim. 💯
Great video! I just switched to Minimax from Luna Dream Machine and it is night and day! Prices are better and videos are 10x better. Only down side is the speed of Minimax.
You are always the first one with the AI video generation news and then everyone else follows 👏🏼 good work!!
Thank you so much! I try to be as fast and accurate as I can!
I just tried it before watching this video and was pleasantly surprised they finally added this feature and omg, the results are superior to anything I've tried before, hope it stays that way, even if it doesn't get better!!
Whatr I have found is Minimax is extremley good at logo graphics, very good at spelling and has produced some good concepts for me without using an image just a good prompt
Been playing with it quite a lot these last few days and was begging for the image to video to release. Tonight is gonna be fun :)
Best advice: stay up late! Gen is pretty slow in peak hours, which makes sense. The old hug of death!
@@TheoreticallyMedia Yeah thanks, makes sense. I did 24 videos yesterday, out of the 24, maybe 12 seconds worth of decent footage (being picky). It will be interesting to see how that changes with image-to-vid. On average, I'd say between 5-7 minutes per generation at various times of the day. I like that it tells you, how many are ahead of you.
You should try landscapes.. Things like buildings, corn fields, cabin in the woods. Things that don’t require a lot of motion and see what you get.
I've had those same issues with Gen 3 img2vid where it completely loses the style. It seems to work better for landscapes. After playing around with minimax img2vid I think it's better than Runway's, but Kling 1.5 is still king. It keeps surprising me with how well it understands the prompt and maintains the same style.
This is an interesting development. Tim, sorry, little off-topic, but I wonder - have you ever done a video on those historical portraits to life like animated video channels? I am going crazy trying to figure out which tools they are using to do that stuff. Would love to experiment with it myself.
I haven't! But I'll put it on the list! I've got something in the neighborhood that I've been back burnering for a bit
@@TheoreticallyMedia Awesome! Can't wait!
I purchased a month of unlimited Runway, I'm looking forward to more control over characters because when it gets to that detailed you will be able to create a whole movie from image prompts
Minimax is in another level. It just needs lip sync and of course a subscription plan to remove the watermark and grant commercial use.
Sub just lightly announced! I've got a community post up on it! looks like we'll still have free, but also a standard at 9 and a unlimited at 94...big jump there!
@@TheoreticallyMedia I'll check it out, Tim. Thanks!
2:18 shots fired
Haha, there is a reason they charge NJ drivers to leave the state!
I welcome you. The video is very good, it helped a lot to navigate. I would like to ask how it can be set in the text description so that the video does not move in several directions? I mean that it should be in a side view and in a fixed position.Thank you.
Hi thanks for the video. With luma you can make sequence shots, add video. In minimax the last frame. When will it be possible to do a three-minute scene where two actors are talking in a room, with different camera angles? Shot, reverse shot, title, etc.?
Unsure, but they’re working pretty hard in the dev department!
I’ve actually got a pretty big Luma deep dive coming up!
@@TheoreticallyMedia Hi I wanted to know from you if luma works because here it doesn't work anymore at one point I wanted to ask you if to make Mira Max work without subscriptions you can create new Google accounts But do you know that they also ask for the phone number to create new Google accounts to make Mira Max work for free? Hi
I love Morph-Fu. And your Jersey cap. As soon as MiniMax went image to vid, I was in the frickin queue. I took a screenie and then prompted a new action and it works so well. I'm so excited!
I KNOW! I love Morph-Fu too! I really think it could be used to create a cool sequence, if you built a story around it. Like a fight scene in a dream! Hmmmm....Might need to ponder that!
Tim, your GPT for MiniMax is the best thing since sliced bread. By far it's the best prompt generator I've worked with. Even the English output gives me near perfect results every clip! It's a shame you can't monetize it. I'm about 140 clips into a large project and almost all of it is usable in some shape or form. Too bad we can't get rid of the watermark though.
And I haven't even touched the image to video yet.
YOU slayed me Tim. Driver from Joisy. OMG. Yep Big Apple knows. Part time at stand up Tim? I'd go see ya.
Great stuff as always.
I'd pay for Minimax, look forward to it. They consistently produce the best text to image and I've tried them all.
Should be mentioned that with the free plan you get a watermark on your videos. And compared with Runway Gen-3 I get worse result, but these are the very first tests, maybe minimax needs different kind of prompting than runway.
Try out the Prompt Generator GPT that I put up. It's in the description!
Wow! This editor is actually much better than kling, luma and runwayfor my usecase; animating cartoon character sprites!
Yup! I think, currently at least, Minimax is your go-to for stylization consistency!
if you start the 3 free days does its stop automatically or do you have to stop it? thanks
Minimax has the image thing now. I hope they will have a PayPal option when the time comes ✌🏾
Well this morning they announced on discord: "Tomorrow, we are launching new paid plans designed to elevate your creative experience! Enjoy perks like unlimited credits, faster generation times, queuing multiple tasks, watermark removal, and priority access to the latest feature". It was a really FUN ride.... we will see what happens tomorrow LOL At this point SORA and Movie Gen are you going to have to be 10x better and free for weeks to catch up to the Kling and Minimax sales market. So that might be a WIN for us lol. THANKS again Tim for the Prompt!! Awesome job! Really glad I found you and subscribed at the start of the Pika era. Keep up the good work!
Tim so out of all of them, Klingon, runway, pika , minimax etc, which one do you think overall is the best? I would love to see a shootout video reciew
I know you didn't ask my opinion, but I think Kling is the best.
So, to be honest, I try to avoid doing those "shootout" and "this killed this" kind of videos. I know those things tend to get big numbers on YT, but I also think they can skew perspective depending on what is being tested.
I feel like all of them excel at certain things-- and if there's one thing I've learned from doing these videos, it's that any one of them are capable of flipping the table with one update.
All that said, you did get me thinking about maybe doing a video showcasing what each one of them is great at. THAT might be an interesting video-- and kind of a positive spin on the whole thing!
Great video!
Any tips on how to make the characters seem to be silent and not talking? Cause most of the time, they start yapping ceaselessly for me haha!
Could you get your connections to beg them for an Image to Video API, needing that urgently from them!
I’ll let them know! I did see an API option on the site- so perhaps soon?
How do you save a face or character and keep a consistent look from one clip to the next? Is anyone putting together short films with this?
01:48 The 6-second number may be rooted in Chinese culture where 6-8-9 are good luck numbers (4 is the worst and always avoided). Funny thing this is a Chinese tool but this new feature is not yet available here in China. Hopefully is in the next hours, I can't wait!
Ohhh, that IS interesting!! and-- makes sense! I wonder if in the states, we'll never get a 13 second generation!
Also, I didn't realize Minimax hasn't released I2V in China-- BUT, you guys DO get Bytedance's Seaweed-- so, you've got a bit of a win there!
cool, i saw some vids w watermarks and some with not. Can they be removed?
Thankyou! Thankyou Thankyou! Tried this out and used the prompts. This is gonna be so fun! Thankyou for your knowledge and time you took to make this video.
I've played with it. It's good with photorealistic stuff, but not so much with cartoon stuff, flat-shaded stuff, hyper-stylized stuff, etc.
The 3d stuff was pretty good, but yeah, that one cyberpunkish girl, I did note that she sort of remained 2Dish? I can see that being something the model isn’t great with
Hi RustyDusty here, thanks for mentioning me.
Hey man! Great work on the final frame trick!! I was so happy to see that!
Ah, thats why its so slowly today… didn’t noticed the small image symbol till yet.
Cool to see, would be helpful to see what prompt you used for each of these examples.
Thanks for the vid 🤗 very useful
I tried image to video today, 30 generations. Honestly got disappointed. Image to video has a lot of deformation. It is good in simple generations
Text to video is much better. But not I want
What were your prompts? I have found it to be really picky with prompts. Try the crazy long prompt addition in the description.
@@TheoreticallyMedia I just write camera zooms in or out. Maybe it has problem with my camera prompt
It gives good results only if u use 16:9 ratio image
Try to use this at the end of each prompt. It works for me:
No ARTIFACTS, SLOW, UGLY, BLURRY, DEFORMED, MULTIPLE LIMBS,
CARTOON, ANIME, PIXELATED, STATIC, FOG, FLAT, UNCLEAR,
DISTORTED, ERROR, STILL, LOW RESOLUTION, OVERSATURATED,
GRAIN, BLUR, MORPHING, WARPING.
Fight scenes might not be perfect but I'm seeing a lot more movement than on other platforms
Totally! and it's getting close-- The characters aren't teleporting all over the place. I'd say give it a few more months, and we might see something moving that dynamically, without the decoherence!
Get weird Tim! Experimental stuff is where the real good work comes from.
Super helpful and informative as always. This is the ai visual media channel I send anyone to who is doing real work.
With all of these advances though, I can't help but wonder...and be amused...that by the time this stuff is 'perfected' ... nobody is going to care about movies anymore. Between quality of direction, attention span, writing, hollywoodism being so prevalent, and fatigue, its possible we end up where people just want ... physical art and jewelry lol. We'll see what we end up with as corps bleed us dry.
This is an amazing vid gen! first shot got me AMAZING results! thank you so much for your efforts and sharing this!
Oh man, you just got me thinking about Minimax 2.0!
@@TheoreticallyMedia oh nice!!
Thanks for the long prompt Tim, it actually works.
Finally accessible from Japan.
I appreciate your humor, Tim. 😂
A true connoisseur of dad humor I see! Haha, excellent!
@@TheoreticallyMedia I catch every one. Adding a little byte of AI humor is good for the soul.
O yea btw 7:00 min. In you say that there's a large percent of us that dislikes when you get all experimental, well i for one am not in that # my friend. I in fact enjoy when you get experimental. 😊
Appreciate that! Haha-- I'll still get weird, I can't not!
Lol there's nothing wrong with that my guy. Enjoy!😊@@TheoreticallyMedia
Thank you so much for keeping on point with all the ai-ness. The thing for me with Minimax and the free tier is that you get 100 free tokens a day and that allows you to create 3 videos at 30 tokens a shot with 10 tokens left over that don't roll into the next day 😂. But nonetheless, onward and upward!!!!
Thank you! If you're interested in seeing what advanced AI video generators can do, I’ve made a fully AI-generated 3-minute music video on my channel, along with a 1-minute short created using LTX Studio, for comparison. I’m not sure if I’ll be able to achieve the same results with Minimax, but I’m definitely going to give it a try.
I heard that as of tomorrow MiniMax will cost $95 per month for unlimited generations. Same as Runway Gen 3 Alpha, which does so much more.
Is this still a good deal?
Also, Adobe's Firefly Video is dropping sometime this year, and it's included with the Creative Cloud subscription.
They'll also have a $9/mo plan if you want to test the waters!
Nice!
I've been pretty happy with Runway ML, but Hailuo is ...maybe better?
So far I've only used it to animate an AI-generated home studio that I plan to use with a green screen. I wanted to animate the images on the computer monitors, and other very minor movement of foliage WITHOUT camera movement. Minimax seems better than Runway at this.
So far, characters aren't working as well for me as Runway. When my character's head turns, it warps, and doesn't exactly keep the same features. Also, the teeth look strange.
But still impressive, and I love that I have got a loopable background for a green screen set.
Just love ur videos, u keep it straight forward and humorous at the same time... Nice job my friend... 😅😊
Oh, thanks so much! Yeah-- this stuff is all so crazy, you just have to laugh!
In the midnight hour I cry mo, mo, mo! Mo Minimax! I gotten lots of good stuff from them! Can’t wait to see what it does with my midjourney images!
Comment of the day, right here!
@@TheoreticallyMedia 😂🤣
Wow, Tim... how do you keep up with all these new developments? It's really impressive! Thanks for your continued efforts.
Coffee. Lots and lots of coffee!
"eye wonk" love it!
The worst of the wonk!
@TheoreticallyMedia very true.
can you plz explain how to continue to further shots with same scene, consistent character, and camera moves when we are creating a movie? because all the view gets changed for next shot, and sometimes character as well. Every expert on youtube is showing just 1 shot. Please let us learn to shoot a scene of 6 shots, 3 minutes long, showing 7 friends. Refer to any sample scene from series FRIENDS. Many thanks
very good generation, problem is how do I use that, needs a way to work over multiple clips with character consistency
current fighting generated videos remind me most of some dragon ball z fights when they are moving so fast it looks like multi armed blur , except unfortunately that warbling floaty slomo action of generated videos feels different , if you speed it up tho and drop frames like animating on 2s it can be neat
Great video! Thanks for the heads up! I think the future of AI will be dozens of different #AIVideo sites, each with their own style specialization. We need a series of "real life" LLMs and then one "anime", one "manga", one "cartoon", and so forth.
I want a cartoon ai that's not anime!
@@harrygoldhagen2732 So, I agree. As someone with three comic book scripts burning a hole in a pocket, that would be lovely. However, here's the tricky part. When you take a photograph of a human, there is no way to prevent people from duplicating your work if they had the same model, setting, so forth. But when you draw, your art style is uniquely yours. And the second you "give" the style to AI, you're out of work. It would be wonderful to have such AI, but you would be asking for a great many sacrifices and I'm not sure that's feasible.
@@dirtydevotee Very true! But the frustration of having many scripts -- I have a bunch! -- just sitting there undrawn, makes me want to have some tool to make them faster.
@@harrygoldhagen2732 I would like to offer that there are LoRAs you can pair with either Dezgo or Rubbrband that will get you a consistent art style. 90% of them are free. Inzaniak has a fantastic "Graphic Novel - Style" LoRA for SD 1.5 on Civit. Perhaps that might help you edge a little closer.
Viewers are shown examples of how the image-to-video feature works, noting that it can sometimes skew towards either a cartoonish or realistic style, depending on the input image. Thank you
As usual very Funformative 😁😁😁
That’s awesome! And so much better than Dullformative!
@@TheoreticallyMedia 🤣
Are you only meant to use parts of the community prompt? As if you use it all then you basically have no space to type what you actually want in the video.
you mean from the GPT? Yeah, someimes chatGPT gets stupid and goes over the max character amount. In that case, I'll sometimes do a little editing. But, to be honest: I've run it "as is" and gotten some pretty decent results as well.
It is a bit of a keyword shotgun spray, and with 2000 some odd characters, there is usually enough.
Folks, give a hand to this youtuber. He's spending as much time in the comment section as he his making the videos😂😂🎉🤜👏
*NONCOMMERCIAL*, it appears for now.
Any word on commercial/paid use in the near future? Thanks, Tim! Love you and everyone else.
D-AI-niel san, I will teach you the secret arts of Morph-Fu :)
Jokes aside, I both enjoy seeing the "Morph-Fu" sequences (they are funny) and think it's a REALLY good benchmark for fast and accurate character movements, so keep them coming!
Thanks as always for your videos
why I can't import the image yet in official website ? is that just for the beta version?
Should be everyone now! Yet refreshing the browser
So signing up I get a "3 day trial" which expires Oct. 18. Does this mean that's when it goes to pay?
Looks like image to video of minimax doesn't care about camera movement prompt at all
it works some times
🔥Fantastic video🔥
New Jersey catching strays 😂
Haha- going to have the whole Garden State coming after me soon!
Looks Good
It's pretty solid-- just...y'know, have patience...she's running slow today for sure. Launch day "hug of death" and all...
Thanks dude. subscribing.
So cool this video generator 😊
Thank you for such a great and early video! I wonder if we will ever see a image to image from Flux as well
Black Forest Labs is working on a Video Model!
Cool i will try minimax for sure , nice video
the differences to runway are sensational, especially in the interpolation of the movements, in the pixa-like girl video and the godzilla video. Normally, the movement in all video generators looked far too slow and not natural
Does this only make short videos?
Minimax is greatm especially for free and while paid options will appear, I hope they can keep the free part going as it is. Although, I'd pay for longer vids with consistency.
It's kinda funny hearing you mention that OpenAI strike as that happens to me with Instagram. I sometimes get strikes for 'spam', even though I've never made a single comment. Not one, as I only use it for movie /TV updates and info. I do the appeal thing and I'm back within minutes, so more of an annoyance than a problem. Makes me think it's a bot screwing up.
Ahhh, sorry to hear, but glad I'm not alone! And yeah, 100% suspect some sort of automation issue. I also think on the ChatGPT side, it might get tricked into giving a few keywords on the no-no list-- and considering that GPT has over 10k uses, I think they likely add up after awhile to an automated strike.
It isn't a huge deal, but it always seems to happen right when I need it!
@@TheoreticallyMedia Why is it always "when you need it." It's like an internet version of sod's law. 😁
It is not free, it has the same predatory credit system as they all have and the quality is poor at best...
I NEED HELP IN USING AI TO GENERATE A 6 EPISODE LIMITED SERIES (EACH EPISODE A HALF HOUR OR BIT LONGER). Scripts are completed.
QUESTION. CAN I UPLOAD A COMPLETED SCRIPT TO MINIMAX OR ANY OTHER AI GENERATOR...I HAVE FULLY DEVELOPED CHARACTERS, DIALOGUE, RECOGNIZABLE FOREIGN COUNTRIES LOCATIONS.
AM I BEING PREMATURE? IS AI THATY ADVANCED?
Perfect 👍🏻. I like minimax. It’s amazing
Runwayml: cinematic trailers
Kling AI: Cinematic and Action. Also gaming style graphics.
Minimax: Action scenes with some distortion.
We are almost getting there.
where's Vidu with the consistent characters and actors?😑🤨
covered it a few videos ago, but yeah-- need to circle back. That is 100% Vidu's secret weapon!
Recent update: It's not free anymore
Well f. Probs to pay for images.
🔥Thank you! 🔥 to Theoretically Media for the fantastic content in your video, The Best FREE AI Video Generator Just Got BETTER! Your insights and thorough explanations made it so much easier to understand the updates and capabilities of this tool. I truly appreciate the effort you put into creating such valuable content. It’s incredibly helpful for those of us looking to enhance our video projects with AI technology!🔥Thank you! 🔥
The new SORA
The "old" Sora never even happened! Haha
@@TheoreticallyMedia 😂
I don't think they had it plus the gov is all up in openai
thanks, I didn't even notice that image to video possibility.
yeah, this is going to help a lot with character consistency. So far I am screenshotting the subject of the prior video, then cropping, uploading and reprompting to continue the scene.
You definitely lose some sharpness in minimax but it's motion repertoire is much higher
Doesn't anybody understand what the word FREE means, that means not having to upgrade your account, that means no restrictions OF ANY KIND
It's letting me generate completely for free as many times as I like. All I needed was a Google account for logging in
Nice i will give it a try ,thks
yes, so far so well, but still no nsfw generations possible! Worthless for me. All AI generators need to address this issue in due time.
Thanks for the update, though.
6:24 That’s why these AI video tools need to add inpainting.