I built Bulin 47 AI, a small browser tool that takes one portrait and makes a version of the Bulin 47 freestyle clip.
The result is a 10-second vertical video. The portrait guides the main rapper’s appearance, while a fixed reference video guides the performance and surrounding scene. The reference soundtrack is included.
The main product decision was to keep it to that one scene.
Deciding what the user actually needs to choose
A general video generator needs room for prompts, model selection, reference footage, duration and output settings. For this particular clip, most of those choices can be made ahead of time.
I’ve already prepared the reference clip, prompt and audio. The person using the site mainly needs to decide who should appear in the video.
That makes the upload form smaller, but it also sets a clear boundary. The tool doesn’t offer custom choreography or a library of different performances. If you want a different scene, this isn’t the right tool for it.
The photo and video reference have separate jobs. A clear portrait helps with appearance; the reference supplies the movement. There’s no need to record yourself performing or upload a full-body shot.
A small form still needs to explain the result
The flow is straightforward: watch the reference, sign in with Google, upload a portrait and submit a generation. You can follow its progress in My videos, then download the finished MP4.
The output is fixed at 720p, 10 seconds and a 9:16 aspect ratio. Those details belong near the upload form, because they affect whether the result will work for someone’s intended use.
There are less tidy details to explain too. A face can look convincing in the first frame and change when the person turns. Hands and timing need checking. The model creates new frames, so appearance and motion can vary between generations.
A completed job doesn’t necessarily mean the person will like the result. That matters when explaining credits: confirmed failed generations return the credit, while a completed video with an unsatisfying likeness is a different situation.
Keeping the examples honest
The homepage includes the reference and trend examples so people can see the scene before paying. Those gallery clips come from outside creators, are credited on the site and weren’t generated by this service.
That distinction needs to be visible. A reference clip shows what the tool is aiming for; it doesn’t prove what every generation will look like.
Making a video requires a paid credit. There are no free generations at the moment. Credits are sold in one-time packs, with no subscription or automatic renewal. The current options are on the pricing page.
A few limits worth stating upfront
Photos are kept for up to seven days and finished videos for 30 days. Download any result you want to keep. Get permission before using someone else’s portrait, and label shared clips as AI-generated.
Keeping the scope small gives me a specific workflow to improve: one portrait, one scene, one downloadable result. I’m interested in whether that constraint makes the process easier to understand, and where the form still leaves people guessing.
You can see it at bulin47.online. If you take a look, I’d appreciate feedback on the upload flow and how clearly it explains what you’ll get.
Top comments (0)