WEBVTT 00:00.560 --> 00:01.000 Okay. 00:01.040 --> 00:09.080 We're going to use file client locally to send a request to file and run flux to image generation tool. 00:09.120 --> 00:11.880 So first we need to pip install. 00:12.120 --> 00:15.680 I recommend installing it in an environment if you don't already have that. 00:23.080 --> 00:26.480 And then you need to load the environment as well. 00:26.480 --> 00:30.280 We have load env here from dot env. 00:30.320 --> 00:32.280 And then we just run load dot env. 00:32.520 --> 00:34.960 And when you have a dot env file. 00:35.000 --> 00:38.280 So just a file on your computer that's called dot env. 00:38.320 --> 00:48.400 In this local space you would have a file underscore key in their file key equals the API key that you 00:48.400 --> 00:49.400 have from here. 00:49.440 --> 00:51.520 So if you run that it's going to tell you true. 00:51.520 --> 00:52.880 This loaded the environment. 00:53.040 --> 00:54.960 And then hopefully you should have your API key. 00:55.000 --> 00:59.440 If you don't have an API key you just need to go to this URL here. 00:59.440 --> 01:02.160 And then you should be able to find find your API key. 01:02.200 --> 01:06.000 Once you create an account and once you have a paid account as well. 01:06.520 --> 01:06.800 All right. 01:06.840 --> 01:08.440 So now let's get into the code. 01:08.800 --> 01:11.200 So we want to import file client. 01:11.500 --> 01:16.540 We want to get the Pil package to get an image. 01:16.740 --> 01:21.940 I want to import requests so we can make the request to get the image afterwards. 01:21.940 --> 01:24.940 And then we also want to get this bytesio as well. 01:24.980 --> 01:26.340 This is for dealing with the image. 01:26.860 --> 01:33.500 But the main thing we do with file is we just submit a prompt and then we submit a model as well. 01:33.500 --> 01:40.020 In this case the prompt is a photo of a rhino dressed suit and tie sitting at a table in a bar with 01:40.020 --> 01:44.060 bar stools, award winning photography, and then this name of the photographer. 01:44.580 --> 01:46.460 And we're going to create the handler. 01:46.460 --> 01:49.580 We're going to go to file client and then submit. 01:49.580 --> 01:51.620 And there's a few different methods that I'll show you. 01:52.180 --> 01:54.860 But then this is the simple way to do it. 01:54.900 --> 01:59.740 We just want to get the result and to do the result we just do handler get. 01:59.740 --> 02:01.460 And then we can print the result. 02:02.540 --> 02:05.660 And let me just show you what that comes back with. 02:16.300 --> 02:18.180 You can see we have an image here. 02:19.020 --> 02:21.220 And there's a width or height. 02:21.670 --> 02:27.110 how much time it took, for example, and the prompt and if we click on this should be able to see the 02:27.110 --> 02:27.550 image. 02:27.550 --> 02:27.910 Here we go. 02:27.910 --> 02:29.510 We have a rhino in a suit. 02:30.070 --> 02:37.550 So in order to get at that image we want to do image URL equals result images. 02:37.550 --> 02:42.750 And then you get the first image because it could come back with multiple if you add that argument in 02:42.790 --> 02:48.190 and then you get the URL of that image, and then you just need to make a request to get the image to 02:48.230 --> 02:49.990 download it as bytes. 02:49.990 --> 02:51.990 And then we can open the image like this. 02:51.990 --> 02:54.510 So if I run that again it's going to run the prompt again. 02:54.990 --> 02:57.030 And then it'll just display the image for us. 02:58.470 --> 02:58.830 Here we go. 02:58.830 --> 03:00.070 So we can see it right here. 03:00.070 --> 03:01.070 So this is a new image. 03:01.070 --> 03:06.430 Every time you run it it's something different unless you use the same seed okay. 03:06.470 --> 03:12.510 Now that we have that we want to function for generating images in general, let's just wrap that and 03:12.510 --> 03:14.910 we'll say call this generate image. 03:14.950 --> 03:16.910 We want to take in a prompt and a model. 03:17.150 --> 03:19.310 And then we're going to do the handler. 03:19.790 --> 03:23.470 We're going to take input the model there the arguments will be the prompt. 03:23.470 --> 03:25.990 And then in this case we wanted to change the image size. 03:25.990 --> 03:28.670 We're going to do a portrait image. 03:28.870 --> 03:30.550 We get the result. 03:31.110 --> 03:38.970 And then when I get the URL, get that image, and then we'll display the image and return the URL. 03:39.010 --> 03:42.730 So now if we want to use that we would just say generate image. 03:42.730 --> 03:48.370 And then we just pass in that prompt and we pass in the model as well. 03:48.370 --> 03:49.770 In this case flux dev. 03:51.170 --> 03:52.650 But actually we could change that. 03:52.650 --> 03:55.250 So maybe just try a different model. 03:56.970 --> 03:59.730 Flux now is another model we can use. 04:06.090 --> 04:06.570 Cool. 04:06.610 --> 04:12.530 Now we have a portrait image and we're using the Chanel model which is a little bit faster but less 04:12.530 --> 04:16.370 detailed than the dev model. 04:16.370 --> 04:17.850 So it just depends on which one you use. 04:17.850 --> 04:22.250 It's also much cheaper, so you can try out all the different models on replicate. 04:22.250 --> 04:27.330 They give you this code, but as you can see it's a relatively simple API request. 04:27.370 --> 04:29.490 So it really makes things a lot easier. 04:29.490 --> 04:30.730 I prefer to do this. 04:30.730 --> 04:36.610 I used to run stable diffusion on my local computer, and there's stuff that always seems to go wrong 04:36.970 --> 04:38.370 in the setup with that. 04:38.370 --> 04:42.210 And yeah, I've switched everything over just to this because it's much, much simpler.