Agent Opus Guide: Step-by-Step & Transcript

[00:00] In about 5 minutes, Agent Opus can turn one song into a fully edited video with motion graphics, source footage, and a scripted voiceover. If you don't know how to use it the right way, you'll burn through all your daily credits and have nothing to show for it.

[00:12] So I've spent the last 3 weeks figuring out exactly what works and what doesn't, and I'm going to break down everything you need to know to get usable results on your first try. When you first land in the video section in Agent Opus, the main thing you'll see is a large prompt box sitting in the center of the screen.

[00:27] Below that, you'll see some of the top videos created by the community and browsing through them is a great way to see how different the results can be. But you've got explainers, tag marketing and a bunch of other formats all created inside the same platform.

[00:39] But let's go back to the font box. This is your main control center and it's also where most credits end up being wasted. The platform looks simple, you type something in and hit submit, but what you put in that box controls everything. The script, the visuals, the pacing and the tone.

[00:52] A font like, make a video about AI, gives Asian Opus nothing to work with. So it guesses on everything, and those guesses almost never match what you actually wanted. And that's one generation gone. Every time I write an Asian Opus font, I use a framework called SALT.

[01:06] Subject, Audience, Length, and Tagged Asset. It only takes about 30 seconds, and it gives the AI everything it needs to produce a clean result on the first try. S is your subject, and the more specific, the better. YouTube's new AI Lightness Detection tool, and what it means for creators, gives a clear

[01:20] angle for the script. A is your audience. If you tell Agent Opus who the audience is, it adjusts the vocabulary, the examples, and the pacing to match. 20-35 year old content creators on YouTube and TikTok produces a completely different script than tech executives or a college student.

[01:35] L stands for length, and it tells the AI how long the final video should be. I cap mine at 40 seconds, because that's what I've found works for the content I make, but you can set it to whatever fits your format. and tease your tagged assets. If you upload images or video files,

[01:47] you need to reference them in a prompt using the at mention system and tell Agent Opus exactly where to place them. If you skip this part, the AI either ignores your files or drops them in random spots. So I'll paste in a prompt that follows salt.

[01:59] Now I'll submit this and let the agent start working. For comparison I testing the exact same topic using a simple one line prompt that says make a video about YouTube new AI detection tool so we can see the difference You can see the result on the right is incredibly generic the pacing is completely off and the script does not match what I need at all

[02:16] But to solve problems, the script writer nails the angle on the first try, and the whole video feels like it was made for my specific audience. While we look at this result, you might have noticed a 5-minute timer running on the bottom right. Agent Opus gives you that amount of time to review the script before it automatically confirms it,

[02:32] and this 5 minute window is where you either save or waste your generation. If you let that timer run out, the software moves on. If the script has a bad fact, a strange sentence, or awkward pacing, all of that carries into the final video, and you've just burned a credit on something you'll never post.

[02:46] So when the script pops up, take a moment and actually read it. You can edit it like a document, cut lines, rewrite sections, and fix any facts that don't look right. For this one, the script looks solid, but I want to tighten the opening and cut one line that fills our filler.

[02:59] So I'll make those edits and hit confirm. And looking at the final generated video, the whole thing came together incredibly well. The painting is tight, and the visuals match exactly what I asked for. If Agent Opus finishes building your video, and it's not perfect, you don't have to accept everything or throw it away.

[03:13] If you don't like a certain clip, or an image looks wrong, you can go to the editor tab right here, and replace that part without regenerating the entire video. You can also regenerate specific scenes. If scene 3 looks off and the rest looks good, you can tell Agent Opus to regenerate only

[03:27] that scene and everything else stays the same. But there is a whole other set of options we have not even touched yet that affects this output even more. If you dial these in correctly, you probably won't need the editor at all. You'll see a few settings right below the prompt box, and getting them right before

[03:40] hitting submit is the easiest way to avoid results that look wrong. The first is hook selection. The hook for your video is generated by Agent Opus, and you can still pre-select the style. The second is your AI avatar. Click the avatar button before submitting and you can create a digital version of yourself

[03:55] that Agent Opus places into the video. Upload a photo, then pair it with a voice, either your own clone or one of the preset AI voices. To clone your voice, just upload or record a 30 second sample and Agent Opus will build

[04:07] a voice model from it You only set it up once and the clone voice carries across every future video If you don want an avatar at all you can run the video faceless with just the AI voice narrating over the visuals Faceless videos tend to be more forgiving because there one less

[04:20] thing for the AI to get wrong, which means fewer wasted generations. Third up is captions. You definitely want to toggle these on, as shorts with captions statistically perform much better. And in the source filter tab, there's a YouTube source toggle. When it's on, Agent Opus pulls

[04:34] clips and content from YouTube as part of its research, and when it's off, it sticks to images, news articles, and other non-YouTube sources. If you want more control over where your visuals come from, turning this off keeps things cleaner. It takes about 30 seconds to configure these settings, and they affect every generation

[04:49] you run. Right down below here in the video section, you can see the community projects I mentioned earlier. You can see the exact prompts someone used for any video. If you find a format you like, you can hit recreate and use their project as a starting template.

[05:01] prompts that already produce good results is faster than guessing from scratch. There's one more tab is worth knowing about and that's motion graphics. If I click into it, the setup is incredibly simple. You write a prompt and choose an aspect ratio and to guide the AI how you want your video to look,

[05:15] you just click the reference button at the top. This splits into video references and image references. Video references include things like 3D assembly and exploded annotation and image references have styles like pastel isometric and luxury miniature.

[05:28] You pick a reference, add your own images if you need to, Drop in a style, write your prompt, and generate. I'll paste in a prompt for a quick product breakdown. And just like that, it builds a perfectly clean animated sequence. Seeing this kind of output shows exactly why this tab gives you a direct pass for any content that is

[05:44] more visual and data-driven rather than narrative. Story Mode expands on stylized generation by adding full narrative control, custom art styles, and the ability to sync everything to your own audio. If I go back to the left sidebar and click into Story Mode, the interface changes completely.

[05:59] Instead of this standard font box, I now see a row of visual style presets. Animation gives you a cartoon storytelling look, while economics creates a polished infographic style that fits business content really well. Then you have claymation for a textured soft motion feel while 2D line delivers a clean hand drawn sketch style Each preset changes how the video looks and Agent Opus applies that style to every frame so everything stays visually consistent I am making a short explainer about why there a 3500 mile fence running across Australia that was built to keep

[06:28] out rabbits. Cremation works really well here because the story is strange enough that the textured handmade style makes it more engaging. So I'll select it and paste in my comp. With this preset, Story Mode builds everything with a thick sculpted texture and stop motion pacing in a

[06:42] claymation style. I only need to give it the story beat and the tone and the preset does the rest. I'll hit generate, review the script, confirm it and the system will start building the video. And the result came out really well. The rabbits spreading across the map look great in the clay

[06:56] style. The soft blue and white clay backgrounds contrast nicely with the sharp rocks and fences and the whole thing feels like something that would make you stop scrolling and watch. The presets are great for quick results. The story mode also lets you build completely custom styles.

[07:09] For the style, I can add a reference image to set the mood or describe the exact look I want. I'll go with the prompt. Now I have a style that doesn't exist in any preset library, and I can reuse it across every future Story Mode project. Story Mode can also build an entire video around an audio file you upload.

[07:24] If you have a voiceover, podcast clip, or voice note, you can upload the mp3, and Story Mode will automatically sync the visuals to your audio. Every visual lands exactly when the voiceover hits that specific moment, which means the pacing is locked to your voice rather than relying on a random algorithm.

[07:38] First I'll upload my recorded voiceover about why movie popcorn cost more than the ticket, then apply my custom style and hit generate. And this one came out the cleanest, because story mode syncs directly to my audio, the visuals move in sync with what I'm saying, and the custom style helps the video stand out.

[07:53] Story mode tends to be more efficient with credits than the main mode because there are fewer variables for the AI to get wrong. You are not pulling outside footage or rendering an avatar. It is just your style, your audio, and the animation engine,

[08:05] which means way through a regeneration to get a result you actually like. Once you understand salt and use the script review step, your first try results start getting better fast. Story mode also opens the door to a completely different style of videos.

[08:17] If you want to try aging opus yourself, go ahead and sign up using the link in the description. Thanks for watching, and I'll see you in the next one.