What Prompt Do I Use to Make a Bulin 47 AI Video?

Look for separate support for a character image and a reference video. An image upload does not necessarily mean the tool can also follow motion from footage. A text-only mode can interpret a description of a street performance, but it does not have the original clip's precise timing or camera movement to inspect.

That distinction also matters when reading tutorials. Dreamina's guide describes an inspired freestyle recreation, while Media.io's guide discusses reference-video and prompt workflows. Choose the approach that matches the actual inputs available in your selected mode.

The following example is for a different tool that accepts both inputs. Upload a clear character photo and reference footage you have permission to use, then adapt the wording to that tool's controls. It is an instruction draft, not a tested preset or a promise of a particular output.

Use the person in my uploaded photo as the lead performer in the reference video. Follow the lead performer's gestures, shoulder movements and timing. Keep the person holding the speaker and the surrounding scene as close to the reference as possible. Keep my face, hair and outfit consistent throughout the clip. Use a vertical 9:16 frame. Avoid added text and extra people.

The photo supplies the appearance you want; the video supplies the performance you want followed. Stating those roles keeps the request easier to understand than mixing instructions about a new person, a new location, different choreography and a moving camera into one paragraph. Begin with the change that matters most: the lead performer.

If the interface has an input-assignment control, use it too. Text cannot reliably correct a photo placed in the wrong input slot. For help choosing the identity reference, see our photo guide. A long instruction to preserve identity cannot recover facial detail that the uploaded image does not show.

Do not assume that adding 'keep the original sound' makes a silent generation mode export audio. Check whether the chosen mode preserves source sound, generates new sound or produces silent footage. Those options have different implications for a performance built around a specific spoken rhythm.

If you are combining video and audio afterward, use the matching recording and check the timing at the beginning and end. Even correctly aligned sound does not guarantee accurate mouth movement. Watch the result with audio instead of judging the prompt by a still image or an autoplay preview with sound turned off.

After a first generation, name the main issue: the wrong person changed, the face drifted, the background shifted or the movements no longer followed the reference. Check the inputs and available controls before adding more instructions. Keep revisions focused so you can judge whether the next result addresses that problem.

For our built-in generator, skip the prompt entirely. Upload one photo, choose Original or World Cup, and select 480p or 720p. The result is a 19-second vertical MP4 with original-audio and silent downloads. Our creation walkthrough covers the settings and checkout. Neither workflow guarantees a perfect face, exact gestures or flawless lip sync.