WEBVTT 00:00.160 --> 00:06.320 This new AI video model from the TikTok team called sedan's, just dropped a few days ago. 00:06.440 --> 00:11.040 And as you can see, it's already beating Google's V3 and clink two. 00:11.080 --> 00:16.880 What if you could automate the process of generating videos using sedans with this workflow? 00:16.920 --> 00:23.240 Now the clips I'm going to show you in a sec were generated completely automatically using this agents 00:23.240 --> 00:25.560 and automation workflow you are seeing right now. 00:25.640 --> 00:26.800 So let's watch them now. 00:35.960 --> 00:41.760 Vo three has been the top AI video model since Google dropped it a few weeks ago, and people are using 00:41.760 --> 00:43.280 it to blow up on social media. 00:43.320 --> 00:47.080 There are channels that went from zero to tens of thousands of followers in weeks. 00:47.120 --> 00:48.480 Just look at this example. 00:48.760 --> 00:53.880 This account has over 70,000 subscribers and over 2 million views. 00:53.920 --> 00:55.040 Let's watch it quickly. 01:12.590 --> 01:16.750 Now let's watch the clips I generated in two minutes using the workflow. 01:30.470 --> 01:36.950 Now, what's really cool is that Siddons not only generates better quality videos, it can also be 25 01:36.990 --> 01:38.830 times cheaper than Vo3. 01:39.150 --> 01:46.630 So instead of paying $0.75 per second like you would with Vo3, you are looking at maybe $0.03 per second 01:46.670 --> 01:47.510 with Siddons. 01:47.870 --> 01:48.310 All right. 01:48.350 --> 01:50.470 Now I'm going to walk you through how this setup works. 01:50.590 --> 01:55.350 Now, this might look complex at first, but it's actually pretty simple once you understand the flow. 01:55.510 --> 01:58.020 So first of all, you have four main steps. 01:58.380 --> 02:03.740 Generate ideas, produce clips, generate sounds, and connect clips. 02:04.260 --> 02:07.620 So starting here on the left we have our idea generation. 02:08.020 --> 02:09.940 So this runs on a schedule trigger. 02:10.140 --> 02:13.620 So every day or however often you want it kicks off automatically. 02:13.820 --> 02:16.340 We've got two agents working together here. 02:17.020 --> 02:19.260 The first one generates. 02:19.260 --> 02:24.780 The first one generates unique video ideas based on whatever niche you are targeting. 02:24.780 --> 02:32.100 And based on your prompt, it's creating the full concept, writing viral captions with hashtags, describing 02:32.100 --> 02:37.980 the environment where the video should be shot, and even what sounds should go with it. 02:38.380 --> 02:44.380 Because we generate video clips and audio separately, then the second agent takes that idea and breaks 02:44.380 --> 02:47.700 it down into multiple detailed scene prompts. 02:47.980 --> 02:54.900 So from one idea, we are getting like 10 to 13 different scene variations that we can fit to sedans. 02:55.100 --> 03:01.340 Then the second agent takes that idea and break it down into multiple detailed scene prompts. 03:01.740 --> 03:08.460 So from one idea, we are getting like 10 to 13 different variations that we can fit to sedans. 03:08.660 --> 03:13.260 So our AI video generation model now we move into clip production. 03:13.500 --> 03:15.500 So this is where we connect to sedans. 03:15.820 --> 03:19.700 But we are not going directly to ByteDance. 03:20.260 --> 03:27.780 So the team behind the sedans we are using instead we are using wave speed AI, which is essentially 03:27.820 --> 03:34.940 a platform that bundles together hundreds of AI models and gives you access to them through one simple 03:34.940 --> 03:35.660 interface. 03:35.780 --> 03:42.700 So instead of learning how each individual AI model works, you can just use wave speed and access tons 03:42.700 --> 03:45.540 of different models through one service. 03:46.100 --> 03:51.500 They're actually one of the first to implement Sedans Pro, so you can use the model that just returned 03:51.540 --> 03:52.100 V3. 03:52.540 --> 03:58.370 Then we take this scene prompt and sent them one by one to see the answer through Wave Speeds API and 03:58.370 --> 04:03.370 we are using HTTP requests so the system waits for the clips to generate. 04:03.450 --> 04:10.250 Usually it takes about 2 to 3 minutes, then pulls back the video files using another HTTP request node 04:10.370 --> 04:11.490 to get the results. 04:11.610 --> 04:16.490 So from our original idea, we now have multiple short video clips ready to work with. 04:16.530 --> 04:20.770 Since sedans doesn't generate audio, we need to add sounds separately. 04:20.970 --> 04:25.250 For this, we are using full AI with the audio v2 model. 04:25.530 --> 04:31.410 This basically takes each video clip, analyzes what's happening visually, and layers appropriate sound 04:31.450 --> 04:32.610 effects on top. 04:32.730 --> 04:39.610 And the cost for this is really cheap, like a 10th of a cent per second, but it makes a huge difference 04:39.610 --> 04:40.850 in the final quality. 04:41.330 --> 04:47.250 So we connect to full AI to send the video clips and then get back the clips with the added audio. 04:47.850 --> 04:53.010 And finally we stitch everything together, so the system takes all those individual clips with the 04:53.010 --> 04:58.080 their sounds and uses for AI to sequence them into one longer video. 04:58.840 --> 05:05.200 Usually we end up with about 30s of content, and that flows naturally from scene to scene. 05:05.400 --> 05:11.880 Now, throughout this whole process, everything gets logged into Google Sheets so you can track your 05:11.880 --> 05:18.560 ideas, see which video got produced, and have URLs for all your final videos. 05:19.040 --> 05:23.560 If you want to take it further, you can connect this sheet to another automation that posts directly 05:23.560 --> 05:27.480 to TikTok, Instagram, YouTube, or any social media. 05:27.760 --> 05:33.960 The whole thing costs around $0.93 per video, which is incredibly cheap compared to what you would 05:34.280 --> 05:36.000 pay with other models. 05:36.160 --> 05:40.640 You would need to pay over 20 bucks using Vo three for the same output. 05:40.880 --> 05:42.560 So for the 32nd video. 05:43.040 --> 05:50.240 So imagine having a system like this running 24 over seven for you, which uses cadence which is much, 05:50.240 --> 05:53.550 much cheaper than Vo feel free and completely automated. 05:53.590 --> 05:58.390 All right, now let's actually build our workflow from scratch, node by node. 05:58.830 --> 06:00.830 So I'm going to walk you through each step. 06:00.830 --> 06:03.790 And you can follow me to set this up yourself. 06:04.150 --> 06:06.630 So as always we need to add the first step. 06:07.630 --> 06:10.790 So the trigger node to activate our workflow. 06:11.230 --> 06:13.550 So I'm going to use schedule trigger node. 06:14.750 --> 06:16.270 So let's search for schedule trigger. 06:21.070 --> 06:25.270 And I want this workflow to be activated every day at midnight. 06:25.550 --> 06:28.430 I'm going to leave the default configuration as it is. 06:29.990 --> 06:30.990 All right now. 06:32.550 --> 06:39.350 And now let's add our agent which will come up with new ideas for the videos. 06:39.790 --> 06:42.070 So I'm going to add agent. 06:45.990 --> 06:46.630 All right. 06:47.670 --> 06:49.030 Now let's configure this. 06:50.900 --> 06:53.500 For the source for prompt we want to define below. 06:57.380 --> 06:59.900 We also want to require specific output format. 07:00.420 --> 07:01.980 So I'm going to toggle it. 07:03.500 --> 07:05.540 And now I'm going to paste the prompt. 07:06.820 --> 07:08.660 Let's switch over to expression mode. 07:10.660 --> 07:14.380 Expand it and paste it here. 07:15.300 --> 07:22.980 So the prompt is give me an idea about realistic granite rock is being sliced with a sharp knife. 07:23.780 --> 07:29.460 Have your idea be in this format color material shaped like a random object. 07:29.460 --> 07:31.340 Examples for your inspiration. 07:31.660 --> 07:33.620 Glass shaped like a strawberry. 07:34.100 --> 07:41.500 Dry ice shaped like a snowflake and and diamond shaped like a hat, etc.. 07:42.260 --> 07:52.740 And we also want to use a tool to review the output You can always go to ChatGPT and explain you an 07:52.740 --> 07:54.780 idea for for your workflow. 07:54.780 --> 08:04.060 For example for video generation workflow and ask for provide some expectations and ask for the prompt. 08:05.820 --> 08:07.180 So I'm going to leave it like that. 08:07.460 --> 08:09.300 Now you want to add system message. 08:09.300 --> 08:12.220 So I'm going to add option system message. 08:15.500 --> 08:17.500 And now I'm going to paste the second prompt. 08:18.820 --> 08:20.780 Again let's switch over to expression. 08:25.860 --> 08:27.740 So this one is quite lengthy. 08:28.780 --> 08:29.660 So let's see. 08:30.020 --> 08:37.340 The role is you are an AI designed to generate one immersive realistic idea based on a user provided 08:37.380 --> 08:37.860 topic. 08:37.900 --> 08:43.940 Your output must be formatted as a JSON array, single line and follow all the rules below. 08:43.940 --> 08:44.620 Exactly. 08:45.820 --> 08:49.370 The rules are only return one idea at a time. 08:49.850 --> 08:55.410 The user will provide a key topic, for example box escaping danger in forest. 08:56.370 --> 08:58.410 The idea must be under 13. 08:58.410 --> 09:03.170 Words must describe a worthy action or moment from the topic. 09:03.650 --> 09:06.570 The caption must be short, punchy and viral friendly. 09:08.210 --> 09:09.810 Include one relevant emoji. 09:12.250 --> 09:19.650 Include exactly 12 hashtags in this order for topic, relevant hashtags for all time most popular hashtags, 09:19.650 --> 09:23.770 and for currently trending hashtags based on live research. 09:24.490 --> 09:35.570 So we want our agent to do research and find the best viral topics we can use to generate our videos. 09:35.570 --> 09:38.650 About all hashtags must be lowercase. 09:39.050 --> 09:41.010 Set status to for production. 09:41.010 --> 09:41.690 Always. 09:42.210 --> 09:45.920 I'm going to show you what is the purpose of this sentence in a sec. 09:46.320 --> 09:49.200 The environment must be under 20 words. 09:49.480 --> 09:51.120 Match the action in the idea. 09:51.160 --> 09:51.800 Exactly. 09:52.520 --> 10:00.160 Clearly describe location, key visuals, background statistic, note macro, cinematic, slow motion, 10:00.200 --> 10:00.840 etc.. 10:01.880 --> 10:08.320 Now the sound, because we are going to generate sound separately and then combine them with clips. 10:09.480 --> 10:12.400 So we need to include requirements for sound as well. 10:12.880 --> 10:15.080 So be under 15 words. 10:15.080 --> 10:18.680 Describe the primary sound that makes to happen in this video. 10:18.840 --> 10:22.000 This will be fit to sound model later on. 10:23.320 --> 10:26.600 And for the output format we want to use single line JSON array. 10:27.480 --> 10:28.640 And this is an example. 10:29.360 --> 10:36.320 So caption short viral title with emoji for topic hashtags for all time popular hashtags and for trending 10:36.320 --> 10:37.000 hashtags. 10:37.400 --> 10:43.080 Idea short idea under 13 words environment, stunt and status. 10:44.630 --> 10:54.870 So this prompt will give us a unique idea for our video with a caption that our agent, uh, prepared 10:54.870 --> 10:55.390 for us. 10:55.790 --> 11:03.030 So the core idea and the environment description, uh, a short prompt for the sand and also the production 11:03.030 --> 11:07.910 status so we can track the process of generating videos, uh, in our Google sheet. 11:09.630 --> 11:17.230 In other words, the system prompt defines the agent's role and provides rules to make sure the output 11:17.950 --> 11:18.990 will be as you want. 11:20.310 --> 11:23.510 So you can read through this and tweak it if you like. 11:27.870 --> 11:30.070 So the user prompt is relatively simple. 11:30.630 --> 11:36.670 It's just a command to give us an idea about the specific niche we are trying to to replicate. 11:40.590 --> 11:41.950 Now let's add our chat model. 11:43.740 --> 11:45.980 I'm going to use GPT four. 11:47.740 --> 11:54.140 Make sure your OpenAI account is connected and let's switch to GPT four. 11:55.420 --> 12:06.460 I don't recommend using GPT four because it struggles with splitting scenes for video clips and giving 12:06.500 --> 12:07.620 the proper output. 12:07.820 --> 12:09.260 I will show you what I mean later. 12:09.940 --> 12:11.900 So let's let's use GPT four. 12:11.940 --> 12:12.300 All. 12:16.900 --> 12:17.260 Right. 12:17.260 --> 12:17.700 Great. 12:20.260 --> 12:24.180 Now we don't need memory for the tool. 12:24.860 --> 12:27.060 Let's use think tool. 12:32.540 --> 12:33.060 All right. 12:33.060 --> 12:36.740 And let's leave the tool description as default. 12:57.340 --> 12:59.820 Now the next note is a simple Google Sheet. 12:59.860 --> 13:00.340 Note. 13:02.900 --> 13:09.420 Because we want to save ideas generated by the agent into spreadsheet. 13:12.900 --> 13:14.500 So let's add Google sheet node. 13:19.300 --> 13:21.380 We want to append row in sheet. 13:25.060 --> 13:28.940 Now let's rename this node save idea. 13:29.900 --> 13:35.380 Make sure you are connected to the correct account for the resource. 13:35.740 --> 13:40.250 We have sheet within document for the operation we want to add up and row. 13:41.570 --> 13:44.930 And now I'm going to share with you the sheet template. 13:44.930 --> 13:49.570 So once you save it you will be able to retrieve it here. 13:51.450 --> 13:55.010 So this is the Google sheet we are going to use for this workflow. 13:56.330 --> 14:06.170 We have a few columns ID idea which is generated by our first agent caption production. 14:06.570 --> 14:15.010 You have for production you have done and skip environment prompt, command prompt and final output 14:15.850 --> 14:17.730 which will be populated automatically. 14:17.970 --> 14:21.530 So it will be the link to the final video. 14:23.970 --> 14:32.570 So once you download it from the resource section and save it, you can use it here from list. 14:35.130 --> 14:38.360 Going to see that in a ten Hit one. 14:46.160 --> 14:52.920 And once you select this Google Sheet, it will automatically give you all the columns. 14:53.320 --> 15:00.120 So you have ID, idea, caption, production environment, prompt, sound prompt, and final output. 15:01.560 --> 15:08.360 This Google sheet looks generations, so when this workflow creates the final video the automation will 15:08.360 --> 15:10.440 mark this row. 15:10.440 --> 15:11.960 So production. 15:13.600 --> 15:16.360 So right here as done. 15:17.600 --> 15:21.000 So this Google sheet will be automatically populated. 15:21.240 --> 15:24.040 And you also get the link to the final video. 15:24.320 --> 15:28.680 So you can share it on social media or just download it. 15:29.680 --> 15:36.390 Now each of those columns we are mapping based on what our ideas agent Uh, provided us. 15:38.190 --> 15:41.310 So now let's actually rename this agent to. 15:44.310 --> 15:44.750 Idea. 15:44.750 --> 15:45.470 Generator. 15:50.870 --> 15:51.750 Agent. 15:52.750 --> 15:53.430 Perfect. 15:54.830 --> 15:55.910 Now let's close it. 15:57.030 --> 15:57.630 Save. 16:02.070 --> 16:08.710 And now let's test it to generate ideas so we can map all the values. 16:11.470 --> 16:12.510 To execute workflow. 16:18.270 --> 16:19.310 Let's wait a moment. 16:23.990 --> 16:24.990 Alright, we got some error. 16:24.990 --> 16:26.950 But don't worry, it was expected. 16:29.870 --> 16:34.860 As you can see, at least one value has to be added on the values to send. 16:35.220 --> 16:35.820 So now. 16:37.300 --> 16:44.460 Now as you can see we have an output, but we have all values just within this one object. 16:45.460 --> 16:50.380 So we need to split these values to match them to the columns. 16:51.260 --> 16:53.100 So we are going to use a node. 16:54.860 --> 16:56.420 So let's add another node. 16:57.140 --> 16:58.420 So output parser. 17:01.860 --> 17:05.860 Because we want to return data in a JSON format. 17:10.340 --> 17:10.980 All right. 17:11.620 --> 17:12.700 Now we want to. 17:14.140 --> 17:19.260 Now we need to provide an example of a JSON format from this output. 17:19.700 --> 17:20.620 We want to get. 17:20.940 --> 17:22.380 So I'm going to paste it here. 17:23.340 --> 17:26.780 Going to remove this and paste it here. 17:28.860 --> 17:34.020 So we want our output to be splitted to separate Eight key values. 17:34.980 --> 17:35.940 Show caption. 17:37.780 --> 17:38.340 Idea. 17:38.340 --> 17:39.020 Environment. 17:39.060 --> 17:40.060 Sounds and status. 17:40.620 --> 17:41.140 And now. 17:44.180 --> 17:44.740 Save. 17:46.740 --> 17:49.900 Now let's test the workflow again to execute workflow. 17:51.060 --> 17:51.620 All right. 17:53.900 --> 17:59.780 Now as you can see we have a nice JSON format so we can map the values. 18:00.260 --> 18:03.140 So let's leave it for now for the idea. 18:04.300 --> 18:05.780 Let's grab it and drop it here. 18:07.220 --> 18:08.180 Now caption. 18:10.780 --> 18:14.180 So we are going to pass dynamic variables because the text is green. 18:19.300 --> 18:22.980 Production is from idea generator. 18:23.020 --> 18:25.420 A note included in the output. 18:29.900 --> 18:31.890 Now for production we have status. 18:34.530 --> 18:35.930 For environment prompts. 18:36.650 --> 18:37.810 You have an environment. 18:40.370 --> 18:43.210 And for for some prompt you have sound. 18:48.170 --> 18:52.690 The last one will be populated automatically because this is the final video. 18:54.090 --> 19:02.690 So yes, now each column is properly mapped except for the final output where we will store the final 19:02.690 --> 19:03.250 video. 19:03.290 --> 19:05.450 Once the workflow generates it. 19:05.650 --> 19:10.170 So the ID uses an Excel formula by design. 19:10.690 --> 19:19.130 So when this node runs, the Excel formula gets stored and translates to an ID for our tracking. 19:19.450 --> 19:25.250 So the different values that our idea that our ideas agent will create. 19:25.450 --> 19:26.330 Alright great. 19:27.330 --> 19:28.610 So let's add this formula. 19:29.210 --> 19:30.840 So it's an equal sign. 19:31.240 --> 19:31.640 Row. 19:31.680 --> 19:33.000 Double brackets minus one. 19:33.800 --> 19:42.600 Now when you check the Google sheet and click on any ID, you can see that this is a formula by default. 19:42.880 --> 19:44.160 So you can just copy it. 19:46.680 --> 19:47.880 And paste it right here. 19:50.160 --> 19:50.520 All right. 19:50.520 --> 19:50.720 Great. 19:50.760 --> 19:51.800 Now this note is done. 19:53.400 --> 19:54.160 Let's move on. 19:54.560 --> 19:56.760 Now you can proceed with our second agent. 19:57.720 --> 20:05.680 So this agent will turn those ideas into multiple prompts to then generate three different scenes. 20:05.680 --> 20:12.640 Because we want to generate three separate clips and then stitch them together into one final video. 20:14.080 --> 20:15.320 So let's add an agent. 20:21.520 --> 20:23.920 Now for the source. 20:23.920 --> 20:26.840 For prompt we want to define below. 20:29.110 --> 20:30.430 I'm going to pace the prompt. 20:30.750 --> 20:34.110 So give me three video prompts based on the previous idea. 20:34.430 --> 20:36.710 Use the text tool to review your output. 20:38.830 --> 20:39.270 All right. 20:39.270 --> 20:39.630 Great. 20:40.350 --> 20:42.270 So this is our prompt writer. 20:43.070 --> 20:44.630 So let's rename it. 20:47.030 --> 20:47.550 To. 20:50.150 --> 20:53.590 Prompt writer agent. 20:54.070 --> 20:56.270 And you also require specific output format. 20:59.630 --> 21:01.390 Now let's add system message. 21:04.950 --> 21:08.790 Such that expression expanded I'm going to paste the prompt. 21:10.190 --> 21:20.070 So the role is you are a prompt generation AI specializing in cinematic ASMR style video prompts. 21:20.070 --> 21:26.580 And your task is to generate a multi-scene video sequence that vividly shows a sharp knife actively 21:26.620 --> 21:30.260 cutting through a specific object in a clean, high detail setting. 21:30.580 --> 21:32.540 So your writing must follow this style. 21:32.740 --> 21:35.140 Sharp, precise cinematic realism. 21:35.460 --> 21:40.620 Macro level detail with tight focus on the blade interacting with the object. 21:41.220 --> 21:46.060 The knife must always be in motion, slicing, splitting, or gliding through the material. 21:46.300 --> 21:48.380 Never idly or static. 21:49.980 --> 21:51.660 Camera terms are allowed. 21:51.860 --> 21:59.140 Macro view tight angle over the blade shot, and each scene must contain all of the following expressed 21:59.140 --> 22:00.860 through detailed visual language. 22:01.180 --> 22:03.380 The main object or subject from the idea. 22:03.740 --> 22:06.700 The cutting environment or surface from the environment. 22:06.700 --> 22:13.740 So as you can see from our previous note, idea and environment, the texture, structure and behavior 22:13.740 --> 22:20.820 of the material as as it's being cut, a visible sharp blade actively cutting, and descriptions should 22:20.820 --> 22:25.740 show the physical makeup of the material, how the material responds respond to the blade. 22:25.860 --> 22:28.100 The interaction between the blade and the surface. 22:30.620 --> 22:31.260 The tone. 22:31.860 --> 22:32.340 Clinical. 22:32.340 --> 22:32.860 Visual. 22:33.660 --> 22:36.340 No poetic metaphors, emotion or storytelling. 22:36.620 --> 22:39.460 Avoid fantasy or surreal imagery. 22:39.980 --> 22:43.740 All descriptions must feel physically grounded and logically accurate. 22:44.100 --> 22:44.740 And length. 22:45.220 --> 22:49.740 Each scene must be between 1000 and 2000 characters. 22:50.700 --> 22:53.020 No shallow or repetitive scenes. 22:53.060 --> 22:55.860 Each must be immersive, descriptive and specific. 22:56.260 --> 23:00.100 Is scene should explore a distinct phase of the cutting process. 23:00.140 --> 23:05.180 A different camera perspective or a new behavior of the material under the blade. 23:06.140 --> 23:06.860 And now for the. 23:06.900 --> 23:14.580 From the inputs we want to pass the values from the previous node. 23:14.580 --> 23:15.380 So idea. 23:15.580 --> 23:21.100 So you can simply just remove it and take idea from here. 23:23.930 --> 23:25.650 The symphonic environment and sound. 23:26.570 --> 23:30.410 So here we define the format we want to achieve. 23:31.450 --> 23:32.330 And since. 23:33.130 --> 23:37.770 So that's why we we want to toggle required specific output format. 23:40.170 --> 23:40.970 Alright great. 23:40.970 --> 23:44.450 So as you can see this prompt is very detailed. 23:45.770 --> 23:47.410 You can tweak it. 23:47.410 --> 23:53.730 You can modify it to to your needs depending on what videos you want to generate. 23:54.370 --> 23:59.450 This prompt was prepared specifically for the demo you watched at the beginning of this lesson. 24:01.330 --> 24:02.650 Now let's start our chat model. 24:04.850 --> 24:08.370 So actually we are going to use the same GPT photo. 24:09.010 --> 24:10.770 So just let's connect it to this node. 24:11.970 --> 24:13.010 We don't need memory. 24:15.130 --> 24:17.850 For the tool we want to use node as well. 24:19.490 --> 24:20.610 So let's connect it. 24:22.800 --> 24:23.160 Well, I. 24:23.160 --> 24:23.760 Perfect. 24:26.600 --> 24:27.720 Now for this agent. 24:28.000 --> 24:31.160 We also want to use a structured output parser. 24:31.560 --> 24:32.440 So let's add it. 24:34.760 --> 24:37.520 Because we want to define a specific JSON format. 24:41.080 --> 24:44.000 So how the output from our agent should look like. 24:45.480 --> 24:47.600 So I'm going to paste the JSON example here. 24:50.200 --> 24:54.760 So you have idea environment sand C1, C2 and C3. 24:55.080 --> 25:00.800 Because as I mentioned we want to generate three separate scenes and then stitch them together into 25:00.840 --> 25:01.960 one final video. 25:05.840 --> 25:06.200 All right. 25:06.200 --> 25:06.800 Perfect. 25:09.000 --> 25:11.480 Now let's actually save and test this workflow. 25:29.230 --> 25:30.230 Let's wait a moment. 25:35.670 --> 25:36.430 Let's check. 25:42.030 --> 25:45.230 Now, as you can see, we have nicely formatted output. 25:46.390 --> 25:49.350 If idea environment SAT and three separate scenes. 25:51.910 --> 25:55.630 So the idea is actually gray granite shaped like a donut. 25:56.190 --> 25:56.750 All right. 25:57.990 --> 25:59.470 You have a prompt for our sound. 25:59.510 --> 26:04.590 So sharp slicing sound with Rocky crunch the environment. 26:04.590 --> 26:08.790 You have a close up of sharp knife slicing through ground in workshop lighting. 26:12.470 --> 26:16.430 And for the scene, for the first scene, we have the camera zooms in. 26:17.630 --> 26:24.540 For the second scene we have a close up tight angle captures the right and over the blade shot highlights. 26:25.020 --> 26:26.460 Okay, perfect. 26:27.460 --> 26:29.860 Now let's check our Google Sheet. 26:30.540 --> 26:33.180 As you can see, one more idea has been added. 26:33.620 --> 26:38.300 We have ID, we have a gray granite shaped like a donut. 26:39.700 --> 26:40.700 So this is the new one. 26:43.260 --> 26:46.180 You have all the values stored in our Google sheet. 26:47.380 --> 26:48.460 So just to clarify. 26:50.820 --> 26:56.060 Scene one, scene two and scene three are all different variations of that one idea. 26:56.060 --> 27:01.460 So we can have different iterations when we pass that along to sedans. 27:02.380 --> 27:06.940 So the prompts are kind of similar to our first agent. 27:06.940 --> 27:10.460 So we have a user prompt which is very simple. 27:11.420 --> 27:18.700 Asking for free video prompts based on the previous idea we guide, we guided On what should do, and 27:18.700 --> 27:24.060 we provide instructions around agents, rules and role. 27:24.780 --> 27:26.740 So how the prompt should look like. 27:28.940 --> 27:29.380 All right. 27:29.380 --> 27:32.020 So now we've got our free prompts ready. 27:32.700 --> 27:36.940 And we need to actually generate the video clips. 27:37.300 --> 27:43.220 So the first thing we need to do might look a bit technical but stick with me. 27:43.500 --> 27:45.740 It's a pretty straightforward. 27:45.980 --> 27:53.580 Right now our free prompts are bundled together as one package in N810, so we need to separate them 27:53.860 --> 27:56.260 so we can work with each one individually. 27:56.860 --> 28:03.460 So we need to add a code node and paste the simple code snippet in JavaScript. 28:04.020 --> 28:05.660 So I'm going to paste the code node. 28:05.900 --> 28:09.980 So basically this is our scene extractor. 28:10.300 --> 28:12.540 So I'm going to rename this node. 28:17.330 --> 28:18.690 This extractor. 28:22.570 --> 28:29.170 So we need to unpack our prompts so sedans can process each scene separately. 28:30.570 --> 28:32.530 So that's what this node does. 28:32.930 --> 28:38.770 So I know that sounds intimidating, but you don't need to write any of this yourself. 28:39.850 --> 28:43.410 You can ask ChatGPT to create this script for you. 28:45.450 --> 28:50.930 Also, if you want to understand what's exactly this code is doing, you can copy and paste it into 28:50.970 --> 28:54.130 ChatGPT and ask it to explain every line. 28:54.770 --> 28:57.850 So the important thing is that after this step you have three. 28:58.210 --> 29:02.290 We have three separate items instead of just one package. 29:05.650 --> 29:06.290 So now. 29:08.890 --> 29:13.050 Just to demonstrate it to you when I pin data. 29:19.560 --> 29:20.760 And execute this node. 29:24.600 --> 29:27.280 I can see that I got three items. 29:27.840 --> 29:28.760 When I open it. 29:32.200 --> 29:33.920 Go to table. 29:34.920 --> 29:42.680 So here on the left in the input you can see that, uh, we already have three separate scenes. 29:43.240 --> 29:45.360 So C1, C2 and C3. 29:46.360 --> 29:50.480 And each one is its own field inside a single item. 29:50.760 --> 29:54.280 But in n810, that still counts as just one item. 29:54.720 --> 29:56.560 As one JSON object. 29:57.320 --> 30:04.160 What we need is to turn each scene into its own item, so we can send them one at a time to dance. 30:04.840 --> 30:06.640 So that's why we added this code node. 30:07.160 --> 30:10.280 And that's what this code snippet does. 30:12.160 --> 30:18.310 So it simply pulls out those three scene descriptions and splits them into separate items. 30:18.910 --> 30:21.990 As you can see on the right side in the output. 30:24.830 --> 30:25.270 All right. 30:25.270 --> 30:25.870 Perfect. 30:26.790 --> 30:31.590 Now we can move on to the actual video generation part. 30:33.590 --> 30:37.790 So we are going to use HTTP request node. 30:40.790 --> 30:41.950 HTTP request. 30:43.630 --> 30:45.030 Select this node. 30:48.990 --> 30:50.190 Now for the method. 30:53.030 --> 30:55.710 We want to send a request. 30:56.110 --> 30:59.230 So let's change the let's switch the method to post. 31:08.270 --> 31:11.380 Because we are going to send a request to see dance. 31:11.420 --> 31:12.940 Asking to create our video. 31:13.140 --> 31:13.940 Video clips. 31:16.780 --> 31:21.100 But we are not going directly to see dance. 31:21.900 --> 31:24.140 As I mentioned in the demo. 31:24.540 --> 31:28.260 Instead we are using a service called Wave Spy. 31:31.860 --> 31:39.940 We are going to use this service because they integrated a lot of different AI models into one platform, 31:39.940 --> 31:49.460 and instead of figuring out how to connect up some of the models and using different AI services, we 31:49.460 --> 31:56.140 can just learn how to use AI and we can get access to all of the models. 31:58.140 --> 32:05.700 And they were actually the first to implement Sedans Pro, which is the version that outperformed Google's 32:05.700 --> 32:06.260 V3. 32:07.220 --> 32:10.100 So we are specifically Using there. 32:13.740 --> 32:22.980 So basically the host, a lot of powerful AI models the text to video, image to video and so on. 32:23.300 --> 32:25.340 And you can make API calls to those models. 32:27.740 --> 32:31.980 And they offer good prices compared to the competition on the market. 32:32.220 --> 32:34.780 I guess they are one of the best currently. 32:35.380 --> 32:36.580 So create your account. 32:37.980 --> 32:42.820 Now once you create your account and sign in, let's search for the model you want to use. 32:44.140 --> 32:46.300 So you're going to use ByteDance. 32:46.860 --> 32:56.140 So the team behind the the creators of TikTok see does V1 Pro text to video 480. 32:58.620 --> 32:59.700 So let's select it. 33:05.140 --> 33:14.250 Now as you can see ByteDance sedans 1.0 is the new generation model, outperforming Cliq two with ultra 33:14.290 --> 33:16.850 fast generation and so on. 33:18.330 --> 33:19.810 Now here you have playground. 33:21.010 --> 33:21.730 You can play with it. 33:21.770 --> 33:27.050 You can give it some prompts and check the check if you like the output. 33:27.890 --> 33:30.250 Now your request will cost $0.15. 33:33.490 --> 33:37.330 So for $1 you can run this model approximately six times. 33:38.090 --> 33:40.930 So we are going to check the documentation. 33:40.930 --> 33:42.490 So select API. 33:44.930 --> 33:48.850 And here you will find all the details you need to set this up. 33:50.730 --> 33:52.410 So this is the URL we need. 33:53.570 --> 33:54.570 So let's copy it. 33:57.010 --> 33:59.330 As you can see the request should be post method. 34:04.290 --> 34:06.010 Let's paste the URL here. 34:08.480 --> 34:08.840 All right. 34:08.840 --> 34:09.320 Great. 34:13.360 --> 34:18.760 Now, as you can see, we want to use header for the content type. 34:18.760 --> 34:23.320 You have application slash JSON authorization bearer and the API key. 34:25.280 --> 34:34.720 So basically if you don't know how to use HTTP request node and make API calls to some services, you 34:34.720 --> 34:41.480 can always go to the documentation of a specific service and find out how to set up the connection. 34:42.080 --> 34:44.400 So now let's go back to our workflow. 34:45.560 --> 34:48.920 Let's switch authentication to generic credential type. 34:52.440 --> 34:54.200 We want to use header authentication. 34:56.160 --> 34:57.400 We want to send body. 34:58.720 --> 35:01.840 We want to switch body content type to row. 35:04.990 --> 35:05.550 As you can see. 35:05.550 --> 35:06.190 Data row. 35:10.670 --> 35:11.910 For the content type. 35:14.550 --> 35:16.830 We have application JSON. 35:20.950 --> 35:26.390 Now for the body let's switch over to expression mode and let's expand it. 35:27.230 --> 35:30.670 And I'm going to paste the body parameter. 35:32.510 --> 35:36.390 So here we specify the aspect ratio duration. 35:37.030 --> 35:42.590 As you can see in the documentation the duration you have a prompt. 35:44.710 --> 35:45.630 For the prompt. 35:45.950 --> 35:47.950 We want to pass video team. 35:48.270 --> 35:50.590 So from our prompt writer agent. 35:51.590 --> 35:52.830 So from this node. 35:55.590 --> 35:56.670 We want to pass idea. 35:56.670 --> 35:59.870 So basically I can do it right now. 36:01.190 --> 36:04.500 So we just want to grab this value and paste it here. 36:07.900 --> 36:14.380 For the for explaining to our credence model what happens in the video? 36:14.420 --> 36:15.900 What should happen in the video. 36:16.140 --> 36:18.340 We want to pass JSON description. 36:20.260 --> 36:23.060 And this comes from the scenes extractor. 36:23.060 --> 36:24.260 So the previous node. 36:27.020 --> 36:28.420 So the same as always. 36:28.660 --> 36:32.500 Just need to drag and drop it here. 36:34.260 --> 36:36.220 And the environment. 36:36.220 --> 36:45.540 So where the video is shot we want to pass environment variable from the output from our prompt writer 36:45.540 --> 36:50.060 agent which described for us the environment. 36:52.740 --> 36:53.140 All right. 36:53.140 --> 36:53.660 Great. 36:53.940 --> 36:54.740 And that's it. 36:56.420 --> 37:04.340 Now as you can see all these three variables are dynamic because they are green and here we can see 37:04.340 --> 37:05.020 the result. 37:05.420 --> 37:13.260 So the actual description, the actual idea and environment description of the environment here. 37:13.260 --> 37:19.580 So the prompt is dynamic which means it automatically pulls in the team the detailed action from our 37:19.580 --> 37:22.220 agent and the environment description. 37:23.140 --> 37:31.540 And since we are processing three different prompts you can click through and see what we are actually 37:31.540 --> 37:32.220 sending. 37:32.540 --> 37:38.420 So three separate requests to a wave speed AI platform. 37:41.820 --> 37:42.020 Here. 37:42.060 --> 37:42.300 One. 37:42.340 --> 37:43.380 Two three. 37:48.180 --> 37:48.820 Perfect. 37:50.660 --> 37:51.780 Now that's not batching. 37:51.780 --> 37:56.700 So let's add option batching. 37:59.020 --> 38:02.170 Items per batch, so let's set it to one. 38:03.930 --> 38:11.290 So basically what this means is input will be split in batches to throttle requests. 38:12.850 --> 38:20.450 So we sent one prompt at a time with a short three second pause between each one. 38:20.850 --> 38:24.370 So let's change it to 3000. 38:25.170 --> 38:26.370 3000 milliseconds. 38:27.490 --> 38:27.810 All right. 38:27.810 --> 38:30.170 Great save. 38:31.450 --> 38:36.570 Let's test our workflow to see if everything works properly. 38:40.290 --> 38:40.690 All right. 38:40.690 --> 38:41.210 Great. 38:45.010 --> 38:45.610 It's all green. 38:45.650 --> 38:46.490 That's a good sign. 38:47.490 --> 38:48.330 Let's open it. 38:51.810 --> 39:02.440 So as you can see, the status code is 200, Which means in the API language that everything worked 39:02.480 --> 39:03.200 perfectly. 39:04.240 --> 39:06.120 And you have three success messages. 39:08.960 --> 39:11.240 And that's because I already set up my credentials. 39:11.560 --> 39:12.520 Now it's your turn. 39:12.520 --> 39:15.440 So go to Header Authentication. 39:16.920 --> 39:18.040 Click on this little pen. 39:23.440 --> 39:25.560 And here we need to provide API keys. 39:26.120 --> 39:28.080 So let's go back to documentation. 39:32.840 --> 39:36.080 As you can see authorization is better and API key. 39:36.560 --> 39:41.080 So we need to type exactly as it is. 39:41.840 --> 39:45.400 So switch over to expression mode. 39:47.200 --> 39:48.080 Expand it. 39:52.360 --> 39:53.200 Remove it. 39:54.080 --> 39:54.880 Type better. 39:58.190 --> 40:01.470 Now go to Wave Speed AI platform. 40:04.710 --> 40:06.150 Go to API keys. 40:10.910 --> 40:15.310 And in order to use this service, you need to top up your account. 40:16.190 --> 40:24.070 And here you can buy credits so you can enter a specific amount for example $2 or buy packages for ten, 40:24.110 --> 40:25.390 50 or $100. 40:26.350 --> 40:32.430 Remember, the cost of each API request is $0.15. 40:33.630 --> 40:40.190 I spent $12 to top up my account and I also got $20 as a gift. 40:40.230 --> 40:41.390 Actually, I don't know why. 40:42.470 --> 40:42.750 Maybe. 40:43.030 --> 40:49.950 Maybe if you buy a $10 package you will get a recharge from admin, but I'm not sure. 40:51.030 --> 40:52.310 Once you top up your account, you can. 40:52.830 --> 40:55.590 Once you top up your account, you can go to API keys. 40:58.470 --> 40:59.870 And create a create key. 41:01.710 --> 41:06.230 Please enter a key name so it does. 41:09.470 --> 41:10.430 Create key. 41:12.630 --> 41:13.870 Now you can copy it. 41:15.950 --> 41:17.350 Go back to your workflow. 41:18.270 --> 41:19.190 Paste it right here. 41:21.310 --> 41:22.310 And then close it. 41:22.750 --> 41:25.470 It's safe and perfect. 41:26.510 --> 41:29.750 Now when you test the workflow you should get success message. 41:30.830 --> 41:31.150 All right. 41:31.150 --> 41:31.790 Brilliant. 41:32.790 --> 41:40.230 The key thing to understand is that at this point, we successfully requested our video clips, but 41:40.230 --> 41:42.870 sedans is still working on creating them. 41:43.070 --> 41:48.150 That's why our next step will be to wait for the generation to complete. 41:48.790 --> 41:52.110 So let's add wait node. 41:56.740 --> 42:00.260 Let's wait for 300 seconds. 42:02.540 --> 42:02.820 Yes. 42:02.820 --> 42:04.300 So here we have to be patient. 42:05.700 --> 42:06.700 Just a quick note. 42:07.940 --> 42:09.500 In the HTTP request node. 42:09.860 --> 42:11.060 Make sure that. 42:13.660 --> 42:17.540 The name is spelled correctly. 42:17.540 --> 42:23.580 So authorization with capital A and for the value also. 42:28.660 --> 42:29.260 Better. 42:31.740 --> 42:39.220 So you have you need to make sure that you get the capitalization right and include the space after 42:39.220 --> 42:39.700 better. 42:40.060 --> 42:44.900 And after that paste your API keys from wave speed. 42:49.740 --> 42:51.260 Otherwise it won't work. 42:52.450 --> 42:53.050 Safe. 42:58.410 --> 43:01.530 And actually now let's rename this node to. 43:03.730 --> 43:04.770 Create clips. 43:10.450 --> 43:14.210 Now let's add another HTTP request node. 43:14.210 --> 43:17.490 But this time to get clips. 43:18.930 --> 43:20.890 So search for HTTP request. 43:26.650 --> 43:29.050 So we are using get method instead of post. 43:29.570 --> 43:34.770 And now we're asking wave speed I to give us the finished clips. 43:36.530 --> 43:39.010 So let's go back to our documentation. 43:45.690 --> 43:47.890 So the model we are using is called. 43:50.440 --> 43:51.240 Let me check. 43:57.800 --> 43:58.520 All right. 44:00.880 --> 44:02.360 It will be quicker. 44:08.760 --> 44:09.840 So again API. 44:11.440 --> 44:14.440 And here we have instruction how to get our clips. 44:16.000 --> 44:18.080 So let's copy this URL. 44:18.480 --> 44:20.760 As you can see request get get method. 44:26.600 --> 44:29.800 Let's go to our Get Clips node. 44:30.920 --> 44:38.720 Let's rename it get Clips method get URL. 44:39.680 --> 44:41.080 Now you have to replace this. 44:43.280 --> 44:45.640 But first let's configure this node. 44:46.440 --> 44:52.710 So we want to Authenticate it by generic credential type. 44:53.990 --> 44:56.350 Generic authentication type will be header. 44:57.870 --> 45:00.150 And you have to do exactly the same thing. 45:01.510 --> 45:04.590 So let's check the documentation. 45:06.110 --> 45:09.150 The header we have authorization beta. 45:10.590 --> 45:11.630 So this is correct. 45:12.190 --> 45:13.710 Now let's add our API key. 45:14.350 --> 45:17.390 So switch over to expression expand it. 45:18.990 --> 45:19.670 Type beta. 45:24.150 --> 45:25.270 Followed by space. 45:27.710 --> 45:30.110 I need to grab my API keys. 45:42.110 --> 45:43.190 And paste it here. 45:44.430 --> 45:44.790 All right. 45:44.790 --> 45:45.350 Great. 45:47.030 --> 45:47.710 Safe. 45:49.430 --> 45:49.950 Safe. 45:55.190 --> 45:58.550 Now let's run our workflow so we can get the value. 46:00.070 --> 46:00.470 All right. 46:00.470 --> 46:03.310 So let's click Save and execute workflow. 46:06.190 --> 46:08.670 Oh actually it's starting from the beginning. 46:09.310 --> 46:09.870 All right. 46:10.550 --> 46:11.910 All right, let's wait a moment. 46:14.870 --> 46:22.950 Actually we have to wait a couple of minutes because I specified here 300 seconds to give our agent 46:23.550 --> 46:25.750 to give our model the time to generate clips. 46:26.270 --> 46:27.150 Now, let's move on. 46:27.190 --> 46:28.430 We got three items. 46:29.230 --> 46:32.950 Store three video clips generated by create clips node. 46:34.030 --> 46:36.190 So by sedans by the sedans model. 46:38.630 --> 46:39.630 So let's open it. 46:39.670 --> 46:40.550 We got an error. 46:44.470 --> 46:49.860 This is because we have to specify the request ID. 46:50.980 --> 46:56.940 That's why we wanted to test the first, to now be able to pass the ID from the root node. 46:57.620 --> 47:01.500 Of course, you can always type it, but I wanted to make it easy for you. 47:02.780 --> 47:05.580 So just simply remove this part. 47:09.260 --> 47:10.900 And grab the ID from here. 47:15.980 --> 47:18.580 Actually first let's switch over to expression. 47:25.620 --> 47:26.260 All right. 47:27.020 --> 47:28.980 And now it should work perfectly fine. 47:29.660 --> 47:30.060 Perfect. 47:30.060 --> 47:33.100 So the next step is to generate sounds. 47:33.100 --> 47:38.860 So we can put an audio layer on the clips we got from sedans. 47:39.300 --> 47:40.380 So let's hit save. 47:40.420 --> 47:42.860 Now let's add another HTTP request node. 47:43.340 --> 47:44.810 So search for HTTP. 47:46.850 --> 47:47.530 All right. 47:48.650 --> 47:50.770 Now we want to switch method to post. 47:53.570 --> 48:00.370 And the service the platform we are going to use to generate sounds is called Fallai. 48:01.250 --> 48:03.050 So go to philly.com. 48:03.730 --> 48:05.450 Create your account and sign in. 48:06.330 --> 48:09.450 So it's kind of similar to wave speed I. 48:10.410 --> 48:12.250 And now when you test it again. 48:15.810 --> 48:17.130 So let's pin the data. 48:21.410 --> 48:22.690 And execute this node. 48:28.010 --> 48:30.770 You can see that we get three outputs. 48:30.770 --> 48:33.530 So each containing an mp4 URL. 48:35.970 --> 48:42.130 You can actually you can actually copy this URL and paste it to your browser to see the result. 48:43.000 --> 48:46.240 So it will download the video clip for you to preview. 48:52.040 --> 48:52.680 Let's see. 49:01.960 --> 49:02.280 All right. 49:02.320 --> 49:02.800 Awesome. 49:03.680 --> 49:04.160 All right. 49:04.200 --> 49:09.360 Now that we have our video clips generated, the next steps are optional. 49:09.360 --> 49:11.760 But they are really complete the workflow. 49:11.760 --> 49:18.280 So we are going to add sound effects and then stitch everything together into one longer video. 49:18.880 --> 49:22.600 So for the sound generation we are using another HTTP request node. 49:30.240 --> 49:34.320 And we are posting a request to create sound effects for our clips. 49:36.760 --> 49:41.510 Now the platform we are going to use is kind of similar to a wave speed I. 49:41.870 --> 49:43.430 It's called fallai. 49:44.550 --> 49:48.310 So just go to Philly.com and create an account and sign in. 49:48.990 --> 49:50.230 Tuesday services. 49:50.230 --> 49:56.510 So to make API calls to the models they host, you also need to, uh, top up your account. 49:57.310 --> 50:01.590 But, um, the API calls are very cheap. 50:03.230 --> 50:10.710 So as you can see, I've spent $10 and generated like 30 videos. 50:11.070 --> 50:13.550 And so far I have only used $0.30. 50:14.270 --> 50:17.910 So the model we are using is called model V2. 50:20.430 --> 50:21.870 So once you top up your account. 50:24.750 --> 50:30.630 Just simply type and then audio v2. 50:30.870 --> 50:32.670 So it's the video to video model. 50:35.030 --> 50:43.070 Now we are going to use documentation to figure out how to set up the HTTP request node. 50:43.830 --> 50:46.110 So switch to the API tab. 50:47.590 --> 50:53.430 Then the table of content have to choose http curl. 50:58.470 --> 50:59.390 Then scroll down. 51:01.190 --> 51:05.870 And here we can check how to submit a request. 51:07.590 --> 51:10.030 So the client API handles the API submit protocol. 51:11.150 --> 51:11.750 All right. 51:12.350 --> 51:15.430 So yeah we have request post URL. 51:15.470 --> 51:16.950 So let's copy this URL. 51:20.270 --> 51:22.710 Let's go back to our workflow. 51:22.750 --> 51:24.430 Let's paste the URL right here. 51:25.470 --> 51:31.110 Let's just rename this node to create Sans for the authentication. 51:32.070 --> 51:34.470 Let's choose generic credential type. 51:36.310 --> 51:37.230 We also have. 51:37.230 --> 51:37.460 Of. 51:40.260 --> 51:40.660 Sorry. 51:40.660 --> 51:42.260 We also have a header. 51:43.020 --> 51:47.100 So for authorization we have a instead of better we have a key. 51:48.220 --> 51:50.180 So let's copy on authorization. 51:54.500 --> 51:56.580 Let's select header authentication. 51:58.540 --> 52:06.100 We don't want to use wave speed AI service for this node because we want to generate sounds not video 52:06.100 --> 52:06.660 clips. 52:08.100 --> 52:12.340 So let's provide credentials we have authorization. 52:13.540 --> 52:17.540 Now value expression expand it. 52:20.780 --> 52:25.100 And type capital key followed by space. 52:26.460 --> 52:28.300 Go to your account. 52:29.220 --> 52:32.060 Go to home API keys. 52:32.740 --> 52:33.420 Add key. 52:35.220 --> 52:35.980 Name it. 52:44.530 --> 52:48.850 Remember to keep it safe because it works like a password. 52:50.090 --> 52:51.170 So create key. 52:54.490 --> 52:55.810 Now you can copy it. 52:56.330 --> 52:57.130 Copy key. 53:01.210 --> 53:02.170 And paste it here. 53:03.250 --> 53:04.010 Perfect. 53:04.370 --> 53:05.090 It's safe. 53:07.410 --> 53:08.770 Now I want to send body. 53:10.530 --> 53:14.290 We want to switch body content type to row. 53:27.890 --> 53:28.530 All right. 53:29.210 --> 53:31.370 Now you have video URL and prompt. 53:32.410 --> 53:33.970 We can also expand it. 53:37.520 --> 53:38.720 For the content type. 53:44.560 --> 53:46.760 We have application JSON. 53:52.200 --> 53:54.520 And now let's create our body. 53:55.960 --> 53:59.320 So I'm going to switch over to expression expand it. 53:59.320 --> 54:01.200 And I'm going to paste the body right here. 54:01.720 --> 54:07.360 So the body looks like that as in our documentation we have a prompt and video URL. 54:11.640 --> 54:17.440 So for the prompt I've got ASMR is soothing sound effects. 54:18.000 --> 54:22.440 The prompt our prompt writer agent prepared for us. 54:25.120 --> 54:30.440 So you can simply just go to Prompt Writer agent. 54:36.190 --> 54:36.830 And. 54:40.190 --> 54:42.790 Simply drop it here. 54:44.870 --> 54:55.430 Duration and video URL from our Get Clips node within the output object. 54:59.830 --> 55:00.390 Perfect. 55:02.070 --> 55:08.390 So basically it looks exactly the same as in the documentation because we have a prompt and video URL, 55:08.430 --> 55:09.870 but we also added duration. 55:12.430 --> 55:15.270 As you can see the prompt and URL. 55:18.590 --> 55:19.070 Awesome. 55:20.430 --> 55:21.430 Now let's hit save. 55:23.070 --> 55:24.550 Let's also add batching. 55:24.550 --> 55:24.990 So. 55:27.790 --> 55:29.470 I'm going to add batching. 55:31.630 --> 55:36.230 Items per batch one and 2000. 55:38.430 --> 55:38.790 All right. 55:38.790 --> 55:39.310 Great. 55:40.310 --> 55:41.950 Now let's add another weight node. 55:45.790 --> 55:47.550 So let's wait. 55:49.870 --> 55:56.110 Let's wait about 60s for our model to generate sounds for us. 56:00.470 --> 56:04.510 And now let's get those generated sounds. 56:05.870 --> 56:07.710 So another HTTP request node. 56:14.950 --> 56:18.510 We get sounds method get. 56:19.870 --> 56:21.750 Now let's copy our URL. 56:23.950 --> 56:26.590 Now let's grab the URL from the documentation. 56:33.580 --> 56:34.420 Pass it here. 56:38.580 --> 56:39.500 Let's add. 56:42.380 --> 56:42.740 Okay. 56:42.780 --> 56:43.220 Forward. 56:43.220 --> 56:43.820 Slash. 56:46.500 --> 56:47.300 Request. 56:49.900 --> 56:50.780 Forward slash. 56:51.260 --> 56:56.380 And here you have to pass a request ID from the previous node from the. 56:57.340 --> 57:02.860 So from the wait node after we run this workflow and get it and get it from file I. 57:04.660 --> 57:05.620 So let's hit save. 57:07.500 --> 57:09.300 And now let's actually create songs. 57:10.060 --> 57:11.500 So let's pin the data. 57:15.980 --> 57:18.100 And let's execute this node. 57:21.820 --> 57:24.700 So now we're gonna wait 60s. 57:25.860 --> 57:26.300 All right. 57:26.300 --> 57:27.180 We got an error. 57:27.780 --> 57:28.650 Let's fix it. 57:30.370 --> 57:31.610 Some method not allowed. 57:32.850 --> 57:35.290 Please check you are using the right HTTP method. 57:35.330 --> 57:36.410 Item zero. 57:39.090 --> 57:44.450 So now we are able to grab this request id and drop it here. 57:45.770 --> 57:47.930 Let's switch over to expression. 57:51.130 --> 57:51.890 Perfect. 57:53.810 --> 57:55.170 Now authentication. 57:55.170 --> 57:55.570 Generic. 57:55.570 --> 57:56.530 Credential type. 57:59.010 --> 57:59.730 Question. 58:03.050 --> 58:04.210 Authorization. 58:05.170 --> 58:05.890 Expression. 58:10.450 --> 58:10.850 Okay. 58:13.010 --> 58:14.530 And let's grab our API key. 58:22.610 --> 58:23.690 I'm going to create a new one. 58:35.520 --> 58:36.480 And paste it here. 58:38.320 --> 58:40.200 And remember, it has to be exactly like that. 58:40.200 --> 58:40.600 So. 58:42.960 --> 58:45.320 Key followed by our API key. 58:46.880 --> 58:49.200 Oh, actually, I didn't copy it. 58:49.640 --> 58:50.440 Copy key. 58:50.600 --> 58:51.360 Okay, copy. 58:52.040 --> 58:56.000 All right, let's fix it like that. 58:56.960 --> 58:57.600 Save. 58:59.920 --> 59:00.640 Save. 59:02.320 --> 59:04.160 And let's execute it again. 59:06.000 --> 59:07.160 Now we got an error. 59:18.040 --> 59:18.280 Aha! 59:18.320 --> 59:21.400 Because I have to choose my credentials here. 59:23.040 --> 59:23.800 Perfect. 59:28.150 --> 59:28.910 Let's try again. 59:32.670 --> 59:33.070 All right. 59:33.070 --> 59:34.070 I know what's wrong. 59:34.110 --> 59:36.110 The request is still in the queue. 59:36.550 --> 59:40.310 So basically, the sons are not yet are not ready yet. 59:40.910 --> 59:42.790 So the I. 59:43.790 --> 59:46.910 So the API can send us the result. 59:47.310 --> 59:53.310 So to fix it, we just need to wait a bit longer or check the status first. 59:55.390 --> 59:57.510 Now let's actually restart the workflow. 01:00:01.670 --> 01:00:02.830 Let's wait a few minutes. 01:00:16.750 --> 01:00:20.790 Now the new idea has been added to our Google sheet. 01:00:22.030 --> 01:00:22.430 All right. 01:00:22.470 --> 01:00:25.070 So we have three new clips. 01:00:26.390 --> 01:00:27.910 The message is success. 01:00:30.150 --> 01:00:30.870 All right. 01:00:33.430 --> 01:00:34.790 Now let's wait a few minutes. 01:00:35.750 --> 01:00:35.990 Yes. 01:00:35.990 --> 01:00:40.190 So basically there are two ways to avoid getting this error in the future. 01:00:40.510 --> 01:00:44.590 You can increase the waiting time, for example, to. 01:00:48.310 --> 01:00:51.950 100 80s to give. 01:00:54.270 --> 01:00:57.510 More time to for generating the sense for us. 01:00:58.910 --> 01:01:04.590 Or another way is to add another HTTP request node. 01:01:08.310 --> 01:01:10.030 Before we getting the sense. 01:01:17.270 --> 01:01:18.390 To check the status. 01:01:19.750 --> 01:01:30.820 So you can use the get method to get the status from the AI platform by placing this URL field, and 01:01:30.820 --> 01:01:36.460 then we can use if node to add additional logic. 01:01:36.460 --> 01:01:40.380 So if the status is completed only then move to the next step. 01:01:45.860 --> 01:01:47.300 And then you would need to. 01:01:48.420 --> 01:01:51.860 And here you would need to specify the value. 01:01:51.860 --> 01:01:53.180 So this is our status. 01:01:53.940 --> 01:01:56.980 So JSON dot status is equal to. 01:01:59.980 --> 01:02:00.780 Completed. 01:02:08.940 --> 01:02:13.340 But we are going to stick with the simpler method to not make it more complex. 01:02:14.860 --> 01:02:21.330 I wanted to demonstrate to you how you can fix waiting times errors for the future. 01:02:25.330 --> 01:02:27.690 I'm going to delete these notes. 01:02:30.930 --> 01:02:34.850 And let's try it for our sons. 01:02:36.530 --> 01:02:43.650 Also, I did a test and I realized that I missed a closing bracket in the body in the Create Scripts 01:02:44.330 --> 01:02:44.810 node. 01:02:44.810 --> 01:02:47.610 So right here. 01:02:49.250 --> 01:02:55.010 So it's really important to double check that you don't miss, for example, any bracket when you compare 01:02:55.010 --> 01:02:59.370 the format to the format provided in the documentation. 01:03:00.490 --> 01:03:08.010 So in that case with the documentation on the fly platform, it's a good practice to double check the 01:03:08.010 --> 01:03:12.930 format to the format provided in the documentation of any API and any service. 01:03:17.930 --> 01:03:18.490 Perfect. 01:03:18.530 --> 01:03:21.200 Now when I open the Jetsons node. 01:03:25.160 --> 01:03:28.120 I can see that I've got three items here as the output. 01:03:28.280 --> 01:03:29.160 So that's great. 01:03:30.720 --> 01:03:39.960 And now the important part is that we are passing the description from our original prompt agent, the 01:03:39.960 --> 01:03:41.600 ten second duration. 01:03:41.640 --> 01:03:46.840 And most importantly, the video URL from our previous seasons generation. 01:03:48.920 --> 01:03:55.720 So since we are processing three clips, an audio model will add the sounds to each one. 01:03:56.520 --> 01:03:58.240 That's why we have three items here. 01:03:59.200 --> 01:04:02.960 And when this runs you will see the requests are in the queue. 01:04:03.480 --> 01:04:09.000 And audio model starts working on adding sounds to our clips. 01:04:09.920 --> 01:04:15.520 That's why we add another node here for 60s. 01:04:15.720 --> 01:04:23.390 But I increase the time to 100 and 80s to give a model time to process everything. 01:04:26.590 --> 01:04:30.790 And then we have another Get request to retrieve our clips. 01:04:30.950 --> 01:04:33.190 Now, with sand layers added. 01:04:33.630 --> 01:04:37.830 And each of our three items now contains a video with sound effects. 01:04:39.390 --> 01:04:45.590 Now you can copy this URLs and and preview them in your browser, and you should hear the sound effects 01:04:47.590 --> 01:04:49.150 on top of our original clips. 01:04:49.670 --> 01:04:53.230 Just a quick note regarding the audio model. 01:04:54.350 --> 01:05:03.830 The best part about this model is it costs only about 0.001 per second, which is basically a 10th of 01:05:03.870 --> 01:05:04.470 a cent. 01:05:04.710 --> 01:05:06.630 So it's a very economical for what you get. 01:05:07.350 --> 01:05:07.830 All right. 01:05:07.870 --> 01:05:16.870 Now that we successfully have three clips with sounds, we need to connect to stitch them together into 01:05:16.910 --> 01:05:18.030 one final video. 01:05:18.710 --> 01:05:20.150 So let's add another node. 01:05:21.390 --> 01:05:22.230 This time node. 01:05:22.230 --> 01:05:22.670 Node. 01:05:29.150 --> 01:05:30.950 And what we are going to do here. 01:05:31.470 --> 01:05:39.310 We are going to take this three separate mp4 files and group them into one item. 01:05:39.790 --> 01:05:43.230 So we need them bundled together for the stitching process. 01:05:44.590 --> 01:05:47.070 So I'm going to paste a short piece of code. 01:05:47.590 --> 01:05:53.310 So this code takes our individual video URLs and packages them together. 01:05:54.430 --> 01:05:55.990 Now when you execute this step. 01:05:59.590 --> 01:06:03.030 You can see that we had three separate items. 01:06:03.830 --> 01:06:08.190 And now we have all these videos combined into one item. 01:06:10.390 --> 01:06:13.230 And now we are going to add another HTTP request node. 01:06:14.140 --> 01:06:19.140 This one will be for stitching our videos together using ffmpeg. 01:06:19.300 --> 01:06:20.820 Through file I. 01:06:21.860 --> 01:06:28.380 FFmpeg is a video editing service that lets us do automatic video editing. 01:06:29.260 --> 01:06:33.900 Luckily, Pal-i has an integration with FFmpeg which makes it really easy to use. 01:06:34.500 --> 01:06:36.580 So the cost is very cheap. 01:06:40.180 --> 01:06:41.060 As you can see. 01:06:43.260 --> 01:06:48.900 It's about 0.0002 per compute second. 01:06:53.460 --> 01:06:54.660 So let's set this up. 01:06:55.140 --> 01:07:01.300 First let's actually rename this node to video. 01:07:03.660 --> 01:07:05.780 You are a collector. 01:07:09.860 --> 01:07:15.770 And now let's add another HTTP request node Not to start the stitching process. 01:07:18.370 --> 01:07:19.370 So http. 01:07:24.770 --> 01:07:26.530 Now we want to use post method. 01:07:30.570 --> 01:07:35.530 And to get the URL from Falaise documentation. 01:07:37.970 --> 01:07:41.810 So just search for the book. 01:07:44.810 --> 01:07:46.210 It's this one compose. 01:07:46.810 --> 01:07:48.530 So we did a video model. 01:07:56.090 --> 01:07:56.850 Selected. 01:07:58.170 --> 01:07:59.210 Then go to API. 01:08:04.410 --> 01:08:10.690 And here you can grab the URL and you get the URL from Falaise documentation. 01:08:12.280 --> 01:08:14.600 Authorization stays the same as before. 01:08:15.800 --> 01:08:20.080 So generic credential type header authentication. 01:08:21.520 --> 01:08:22.720 We want to send the body. 01:08:26.320 --> 01:08:36.320 Use raw type for the content type application slash JSON. 01:08:42.720 --> 01:08:46.080 And for the request body we need to add custom instructions. 01:08:46.720 --> 01:08:48.640 So I'm going to switch over to expression. 01:08:50.240 --> 01:08:56.160 Now this comes directly from file documentation on how they use ffmpeg. 01:08:56.560 --> 01:09:03.240 So what we are doing here is we are listing our clips one by one, providing timestamps and duration 01:09:03.240 --> 01:09:08.960 for each, and telling the ffmpeg to stitch them together into one video. 01:09:10.710 --> 01:09:15.870 And now when we run this node, the request goes into queue. 01:09:16.870 --> 01:09:18.670 So let's add another node. 01:09:26.630 --> 01:09:29.910 And let's wait for 60s. 01:09:35.750 --> 01:09:38.830 To let the model stitch our videos together. 01:09:39.350 --> 01:09:41.590 Now let's rename our HTTP request node. 01:09:44.470 --> 01:09:45.430 To sequence. 01:09:47.910 --> 01:09:48.550 Video. 01:09:49.310 --> 01:09:51.790 I forgot about batching, so let's add batching. 01:09:56.710 --> 01:09:59.710 Items for batch one. 01:09:59.950 --> 01:10:03.390 Batch interval thousand. 01:10:09.190 --> 01:10:09.990 Let's hit save. 01:10:11.430 --> 01:10:15.430 And now we need one more get request to retrieve our final video. 01:10:16.390 --> 01:10:25.230 Now, before you give it a test to get the request ID for the next step, make sure you are selecting 01:10:25.230 --> 01:10:30.270 the correct credentials in the authentication. 01:10:30.950 --> 01:10:38.270 So make sure you are using the credentials for connecting to our AI service, not the AI platform. 01:10:39.510 --> 01:10:44.870 All right, now we need one more Get request to retrieve our final video. 01:10:45.870 --> 01:10:48.830 So let's add another HTTP request node. 01:11:01.190 --> 01:11:02.790 Let's switch method to post. 01:11:06.540 --> 01:11:09.140 I'm going to paste the URL from documentation. 01:11:09.780 --> 01:11:11.180 Let me switch over to schema. 01:11:11.980 --> 01:11:17.300 You can grab the request ID from the right node and just drop it here. 01:11:24.140 --> 01:11:24.700 All right. 01:11:25.740 --> 01:11:33.180 So we are using the API URL with the specific request ID that they gave us from the previous step. 01:11:33.180 --> 01:11:36.060 So we can track which video is ours. 01:11:36.780 --> 01:11:40.740 And when this process completes we get our final video URL. 01:11:41.380 --> 01:11:43.740 So just let's add authentication again. 01:11:48.260 --> 01:11:48.820 Add the. 01:11:52.140 --> 01:11:52.700 Whole I. 01:11:57.620 --> 01:11:58.300 Like that. 01:11:59.500 --> 01:12:00.820 Now let's execute this step. 01:12:01.780 --> 01:12:08.290 So we got our final video URL And you can copy it and paste it into a browser. 01:12:09.170 --> 01:12:18.290 It downloads our complete 32nd video with all clips properly stitched together and all the sounds included. 01:12:20.370 --> 01:12:20.730 All right. 01:12:20.770 --> 01:12:21.250 Great. 01:12:21.290 --> 01:12:23.130 Now we have our final video ready. 01:12:23.490 --> 01:12:30.970 So the final step is to properly log that URL in our Google Sheet node. 01:12:31.490 --> 01:12:39.410 So let's add actually first let's rename this node to get final video. 01:12:41.130 --> 01:12:43.490 So now let's add one more Google Sheet node. 01:12:48.250 --> 01:12:49.290 And we want to. 01:12:51.450 --> 01:12:52.770 Update the sheet. 01:12:55.890 --> 01:12:58.170 Make sure you are connected to the correct account. 01:13:03.160 --> 01:13:07.160 Let's select our sedans sheet sheet1. 01:13:09.720 --> 01:13:13.480 The column to match all we want to select idea. 01:13:15.440 --> 01:13:18.520 So this connects to the same document we have been using. 01:13:18.960 --> 01:13:23.480 And what it does is find the specific row we need to update in our Google sheet. 01:13:26.080 --> 01:13:30.640 Remember that row we created at the beginning with our idea? 01:13:35.480 --> 01:13:44.080 We are going to find that exact row and mark the production column as done right. 01:13:45.640 --> 01:13:52.040 And most importantly, we are logging our final video URL into this Google sheet. 01:13:52.040 --> 01:13:55.760 So you have a proper tracking of where to find this video. 01:13:58.640 --> 01:14:02.270 And now let's grab the idea from the Safe Idea node. 01:14:04.870 --> 01:14:06.230 And drop it into this field. 01:14:07.030 --> 01:14:10.030 So we are using the safe idea from the. 01:14:12.670 --> 01:14:13.550 From this node. 01:14:19.750 --> 01:14:22.230 And now we need to replace the item. 01:14:24.550 --> 01:14:26.150 With first. 01:14:28.630 --> 01:14:29.630 And run brackets. 01:14:31.950 --> 01:14:39.190 Because when nodes run in different branches of our workflow and it can't automatically match which 01:14:39.190 --> 01:14:41.150 data belongs to together. 01:14:43.950 --> 01:14:44.350 Yes. 01:14:44.350 --> 01:14:50.950 So we want to tell anything to just to grab the first item from that node. 01:14:55.550 --> 01:14:56.030 All right. 01:14:56.030 --> 01:14:58.830 So now we have a dynamic variable here. 01:15:01.510 --> 01:15:02.190 Perfect. 01:15:03.390 --> 01:15:04.670 Now let's execute the step. 01:15:09.390 --> 01:15:10.070 All right. 01:15:12.110 --> 01:15:13.710 And now let's check our Google sheet. 01:15:17.910 --> 01:15:18.470 Awesome. 01:15:20.030 --> 01:15:20.390 All right. 01:15:20.390 --> 01:15:20.590 Great. 01:15:20.590 --> 01:15:21.350 We are done. 01:15:21.350 --> 01:15:24.270 And if we set this workflow to run every day. 01:15:27.790 --> 01:15:34.470 By using this scheduled trigger, it will create videos for us automatically and log them into this 01:15:34.470 --> 01:15:36.670 Google sheet one by one. 01:15:38.590 --> 01:15:43.070 And now you can either publish this yourself, or you can expand this system. 01:15:43.070 --> 01:15:47.590 To take it a step further, you can connect this Google sheet to. 01:15:48.070 --> 01:15:55.430 You can post these videos on social media platforms such as TikTok, Instagram Reels, YouTube shorts. 01:15:56.390 --> 01:16:06.460 Now to conclude to create 32nd video using CD-Rs and this automation, it was only about $0.09. 01:16:06.580 --> 01:16:12.980 Compare that to Vo3, which would have cost us about $20. 01:16:12.980 --> 01:16:14.340 So that's the massive difference. 01:16:14.940 --> 01:16:18.060 And adding the Suns is also very cheap. 01:16:18.340 --> 01:16:21.140 Only $0.03 for the final video. 01:16:22.180 --> 01:16:25.980 And then stitching the clips together is less than a cent. 01:16:26.580 --> 01:16:33.380 So in total we are looking at $0.93 to create a complete 32nd video automatically. 01:16:33.780 --> 01:16:35.100 So that's pretty incredible. 01:16:35.820 --> 01:16:38.260 All right, I hope you learned something new in this tutorial. 01:16:38.540 --> 01:16:42.220 Remember you can download the template from the resources section of this lesson. 01:16:42.500 --> 01:16:47.740 When you are building this solution from scratch and face any issues or get any errors. 01:16:48.020 --> 01:16:53.980 You can compare your configurations to the pre-made template and always you can ask questions in the 01:16:53.980 --> 01:16:54.980 Q&A section. 01:16:55.980 --> 01:16:58.060 Thank you for bearing with me and see you in the next one.