I wanted a system that could research TikTok and YouTube for promising video ideas, work with footage, make cuts, edit a finished video, and help get it onto YouTube. That project became RankStudio.
The idea sounds simple in one sentence. In an app, it turns into a chain of separate jobs. Research produces references. The references shape a plan. The plan needs real media. The media needs narration, timing, and an edit. The finished render then needs a review and a decision about publishing.
RankStudio gave me a place to connect those stages and keep track of what happened between them.
Starting with a direction
The desktop workflow starts with a direction and the number of drafts I want. The first idea it generates isn't treated as the finished product. It builds drafts I can review, edit, approve, or delete.
YouTube search and TikTok research help me find subjects and formats worth considering. The app can use the structure of an existing short as a reference: how it opens the subject, how the idea develops, and how it pays off. That reference becomes one input to the plan for a new draft.
The distinction that matters here is between a reference and something I can actually use. Finding a popular post can show me a format. It doesn't make that post's footage something the editor can use. RankStudio keeps records of imported assets and requires a reuse basis before it treats source material as footage it can render.
Different kinds of drafts need different paths
The app now has several ways to create a draft. Clip-based drafts use discovery and cleared footage. Animal-information drafts use subject research and supporting media. Explicit joke or meme directions can use generated illustrations and original copy.
That matters because an animal explainer and a visual joke don't need the same production process. The animal workflow includes research sources and citations. The illustrated joke workflow handles the writing and the images differently.
Every planned scene still needs a real asset. When one is missing, the draft fails visibly instead of turning into a video full of unrelated filler. Media availability becomes part of the planning problem.
Assembling the actual video
Once the plan and the assets exist, the edit has to become a real file. RankStudio uses FFmpeg to combine video segments or still images with narration and timed on-screen text. Scene lengths have to fit the spoken material, or the edit won't have a usable rhythm.
The app stores the progress of drafts and renders, so the interface can show where a long operation stands. Researching, planning, rendering, and uploading aren't one button press. A queue keeps those steps understandable and lets several drafts run without mixing up their results.
There's also a reusable media library. An asset or a useful analysis result stays available for later, so I don't start from scratch every time. That gives generation a memory: the system can remember what it already has.
Reviewing before publication
The finished render is what I approve. The desktop workflow uploads the approved result privately to my connected YouTube channel. That's a specific step, and it's more accurate than calling the whole system an unattended public channel.
The review screen lets me look at a draft, change the text, ask for a different take, or reject it. Publishing happens at the end, after the content is visible and a person has made the final call.
The project also has an earlier cloud editor and cloud render components. Over time, the desktop workflow and a Raspberry Pi server role became part of the design. The Pi was meant to keep queues, research, browser sessions, and media available while the laptop was off.
Those versions are part of how the project evolved. They don't prove that every component was running at the same time, or that every publishing path has been tested end to end.
What the project represents
RankStudio is one of my bigger experiments in building a complete AI-assisted system. It covers discovery, structured plans, asset records, queues, editing, rendering, review, and upload.
I don't have a confirmed revenue number or an audience-growth claim for it. The software story is the workflow: I take a direction and move it through enough explicit stages that I can see the results, review them, and work with them.
That's the part I'd show in a portfolio. A video generator needs a lot more around it before it's a tool you can actually run.