Video creation guide
From generated video candidates to a reviewed final export
After generation, select usable takes yourself, assemble them in storyboard order, then review sound, subtitles and the complete sequence before exporting. This is a general workflow, not a product tutorial. Keeping candidates makes comparison and revision easier, but cannot guarantee matching characters, actions or styles across shots. You still need to watch the actual footage.
File candidates under their shots without replacing earlier takes
Label each candidate with a shot ID, version, source and generation date, and keep its references and description nearby. Example: S03-v01 and S03-v02 are two candidates for the same shot, not two additional shots. Separate candidates still under review from selected takes so that a growing folder remains understandable without repeatedly renaming its contents.
Watch each candidate from beginning to end. Check the subject, action, visual defects, duration and fit with neighboring shots. Leave a short selection note, such as “complete action; trim the final moment.” The most attractive still frame is not necessarily the most useful clip, and a new generation should not automatically displace a take you have already selected.
Build a rough cut to reveal gaps and pacing problems
Add the selected takes to a timeline in storyboard order while preserving the originals. On the first pass, concentrate on order, in and out points, and obvious gaps rather than transitions. Play the sequence with a temporary narration track. Check whether viewers have time to read the image and whether an action finishes before the next cut.
Note the time and reason for a problem before choosing to trim, replace a take or create a missing shot. If regeneration is necessary, change the condition that matters most and keep the previous version for comparison. Repair gaps in understanding before polishing small details. Adding effects is not a dependable substitute for an action that does not connect across a cut.
Review picture, sound and subtitles separately, then together
Watch once without sound to spot abrupt jumps, black frames, distortions, stray text and changes in direction. Then listen without relying on the picture: is narration complete, and does background sound obscure speech? Review subtitles for names, numbers, line breaks and timing. Pay particular attention to captions that disappear before there is enough time to read them.
Play the complete sequence and turn observations into specific edits rather than a note that it “feels wrong.” Where possible, spot-check text size and sound clarity on different devices. Keep track of the permitted use of reference material, music and voices, retain their sources, and preserve applicable AI-generation disclosures and markings with the work.
Verify the exported files and keep a recoverable version
Before exporting, check aspect ratio, resolution, frame rate, audio and whether subtitles are embedded or delivered separately. Save the timeline version. Even after an export reports success, open the actual file and inspect its beginning, ending, duration, synchronization and subtitles. A successful task status does not prove playback works. Use matching version labels for the video, subtitle file and project copy.
MiseTV is in development around this path from shot candidates and human approval to basic editing, review and export.