WEBVTT 00:00.660 --> 00:03.330 -: All right, let me show you how to do image to video. 00:03.330 --> 00:06.000 So, first, we're gonna generate an image 00:06.000 --> 00:10.230 and I'm using fal.ai, which is an API you can use 00:10.230 --> 00:11.880 that hosts a lot of these models. 00:11.880 --> 00:13.835 I topped up my account with some credit 00:13.835 --> 00:16.470 but it's like pennies per image, right? 00:16.470 --> 00:18.240 You have some text in the image. 00:18.240 --> 00:19.410 You can see it's really good. 00:19.410 --> 00:22.950 I'm gonna use an actual prompt that I use for my business. 00:22.950 --> 00:24.990 So, this is a style. 00:24.990 --> 00:26.760 It kinda creates these cool 00:26.760 --> 00:30.060 '80s, '90s video game aesthetics. 00:30.060 --> 00:32.380 I'm gonna copy this prompt 00:33.660 --> 00:35.943 and I'm gonna put this in here. 00:36.900 --> 00:40.680 And then I just need to put in some few different variables. 00:40.680 --> 00:41.910 Main subject. 00:41.910 --> 00:43.620 So, this is one I wanted to run, 00:43.620 --> 00:44.853 so I'm gonna show you this. 00:44.853 --> 00:47.100 This is a real one I'm gonna use for a campaign. 00:47.100 --> 00:48.510 So, I wanted to create one 00:48.510 --> 00:50.880 that had a man hailing a yellow taxi. 00:50.880 --> 00:54.330 So, I'm just gonna get rid of that variable 00:54.330 --> 00:55.563 and put that in there. 00:57.360 --> 01:01.140 Man hailing a yellow taxi in a specific environment. 01:01.140 --> 01:04.743 So, this is in a sprawling urban landscape. 01:08.670 --> 01:10.530 And if you're interested in how I made this prompt, 01:10.530 --> 01:12.840 I just went back and forth with ChatGPT 01:12.840 --> 01:15.660 until it had the right aesthetic. 01:15.660 --> 01:18.270 And then I had a look at the prompt 01:18.270 --> 01:20.730 and just copied it and templatized it. 01:20.730 --> 01:22.920 Of course, the composition includes a nearby park 01:22.920 --> 01:26.400 with children playing soccer and people having a picnic. 01:26.400 --> 01:30.210 And then finally, there's gonna be a highlighted object, 01:30.210 --> 01:32.610 which is down here, highlighted object, 01:32.610 --> 01:35.103 so it's a woman in a red dress across the street. 01:36.600 --> 01:39.570 All right, so, let me see how this runs. 01:39.570 --> 01:41.133 Hopefully it works well. 01:43.470 --> 01:44.493 Doesn't always work. 01:47.100 --> 01:48.350 Yeah, that's pretty cool. 01:49.440 --> 01:51.030 I like that. 01:51.030 --> 01:54.000 And yeah, it's got the red dress, I guess. 01:54.000 --> 01:55.740 This guy's arms are a little big. 01:55.740 --> 01:57.730 So, I'm going to download that image 01:59.280 --> 02:03.453 and then I'm gonna go to KLING. 02:06.300 --> 02:10.865 And KLING is the video image generation model. 02:10.865 --> 02:12.948 Gonna search for it here. 02:14.520 --> 02:16.383 Gonna go Image to Image, 02:19.072 --> 02:19.905 And... 02:23.945 --> 02:25.195 And KLING v1.6. 02:27.060 --> 02:30.420 That is the newest model at the time of recording. 02:30.420 --> 02:32.640 And you can see it's pretty good quality. 02:32.640 --> 02:37.140 So, what you need to do is upload the image, 02:37.140 --> 02:41.283 in this case, this one here, 02:42.270 --> 02:44.700 and then you're gonna say what you want it to do. 02:44.700 --> 02:48.967 So, I'm gonna say the taxi drives off with the man. 02:51.060 --> 02:56.060 Drives off after the man gets in. 02:56.790 --> 02:58.410 And you only just need to prompt 02:58.410 --> 03:00.060 what you want to happen in the image. 03:00.060 --> 03:02.580 It's gonna take the style from that image. 03:02.580 --> 03:05.190 And just to note here, it's quite expensive. 03:05.190 --> 03:08.190 It's gonna generate, like, a five-second clip, 03:08.190 --> 03:12.480 and it costs $1.10 per second of video. 03:12.480 --> 03:15.150 So, it's gonna cost you 5 cents to run this. 03:15.150 --> 03:17.913 It takes about five minutes, typically. 03:19.050 --> 03:21.060 Okay, and the video came back. 03:21.060 --> 03:24.720 That took 240 seconds to generate. 03:24.720 --> 03:26.760 See, it's not fast, but the results are good. 03:26.760 --> 03:28.533 So, let me show you. 03:29.610 --> 03:30.783 There you go. Look. 03:31.920 --> 03:33.600 Taxi drives off. 03:33.600 --> 03:35.580 He didn't get in, unfortunately, (laughs) 03:35.580 --> 03:37.590 but it's a pretty good animation. 03:37.590 --> 03:38.610 It's seamless, right? 03:38.610 --> 03:40.440 The consistency, the visual consistency 03:40.440 --> 03:41.550 of the characters is great. 03:41.550 --> 03:43.050 Yeah, I really like this model. 03:43.050 --> 03:45.933 It works really well for small animations like this.