How Frame Highlights Recommends Clips from Five Signals
Highlight recommendations combine motion, audio, faces, shot density, and image quality. The result stays a pending clip until someone accepts or rejects it, one by one.
Recommendations consider five signals together
Frame Highlights recommends clips by combining motion, audio, faces, shot density, and image quality. The five are not five buttons that replace one another. They are signals considered in the same recommendation. The landing page at https://ajuno.xibalans.com/image-video-processing/ uses that sentence to say where recommendations come from. The template decides how those signals are emphasized, and files can be uploaded only after the template is confirmed.
On the default preset 人物动感高能, the weight names that can be read are 运动强度, 音频能量, 人脸出现, 镜头切换, and 画质优秀. Shot density in the recommendation sentence corresponds to the cut signal, and the weight label on the page is 镜头切换. The other four presets have their own wording and do not share the default decimals. A recommendation is a clip waiting for review, not a film already published.
The landing page lists this combination as one capability beside batch upload, templates, multi-ratio covers, the library, and task review. The combination does not add a sixth output such as an automatic post. When the five signals disagree, the template's emphasis is what the run was confirmed to prefer. The reviewer still sees a recommendation, not a finished decision.
Motion strength and audio energy
运动强度 is about visual change such as action and amplitude. The default preset sets it at 1.00 and says it prefers full-body movement, range of motion, rhythm, and performance, while requiring the person to stay present. 综合动感高能 widens fast motion to people, matches, vehicles, equipment, games, and the environment. It is not limited to a full-body person. Both presets talk about motion, but the range differs. The default weight cannot stand in for the 综合动感高能 sentence.
音频能量 is about how strong the sound is. On the default preset its weight is 0.65. The 综合动感高能 text places strong audio together with fast motion and dense cuts. 智能均衡, 人物特写, and 画质优秀 are not labeled with that set of decimals. Audio energy taking part in a recommendation does not mean the loudest seconds must be accepted, and it does not mean the sound will produce views.
A match with a loud crowd and a quiet close-up can both be valid uploads. Which signal leads depends on the confirmed preset. Under the default, motion is weighted at 1.00 and audio at 0.65, with people expected to remain in frame. Under 综合动感高能, strong audio is named in the preset sentence itself, without those decimals attached. Reviewers should not average the two descriptions into a new unlabeled preset.
Face presence and shot changes
人脸出现 is weighted 0.35 on the default preset. 人物动感高能 requires the person to remain present. 人物特写 goes further and asks for a clear person, a close view, and an obvious expression. The face signal is therefore tied to whether the content is about a person, but it is still one of five. A weight above zero only means face presence participates in the default strategy. It does not mean every selected clip must be a close-up, and it does not mean 0.35 is a universal hit rate.
镜头切换 corresponds to shot density in the recommendation description. The default weight is 0.55. 综合动感高能 explicitly includes dense cuts. Frequent cuts are not always better. 画质优秀 does not chase intensity and prefers sharpness, stability, and composition, so dense cutting is not its goal. Read the signal next to the current preset's sentence. Do not treat 镜头切换 as a permanent bonus on every template.
A face that appears for a moment in a vehicle shot is not the same request as 人物特写. A calm, stable frame can still score on 画质优秀 while a dense cut pattern is what 综合动感高能 asked for. The signal names explain the ingredients. The preset sentence says which ingredients this batch was confirmed to prefer.
Image quality does not require intensity
画质优秀 is both a signal and the name of a preset. As a weight on the default template it is 0.45, listed with motion, audio, face presence, and shot changes. As a preset, 画质优秀 does not chase intensity and prefers high definition, stability, and good composition. 智能均衡 instead selects high-quality highlights in combination. Its emphasis is combined quality, not a second copy of the phrase about avoiding intensity.
Image quality taking part in the order does not promise that every recommended frame meets a team's own broadcast standard. It means the analysis includes quality information such as sharpness, stability, and composition. Someone still has to accept or reject each clip on the task. If 精彩采纳率 shows a percentage on the workbench, that percentage is a snapshot of the current workspace, not a universal score computed from these weights.
The preset named 画质优秀 can be confirmed for footage where the point is a clear, steady frame. Confirming it does not turn off the other signals in the product's recommendation sentence. It states the priority: not intensity first. The default preset states a different priority and is the only place those five decimals are given.
Recommended clips still need a person to accept or reject
The five signals produce recommendations. The workbench counts recommendations that have not been accepted yet as 待确认片段, and it says 还有待确认片段，可逐个采用或驳回. 立即处理 is the action that opens that review. Rejecting or accepting does not publish the clip to an external platform, and it does not by itself save the clip as a library asset. Entering the library still requires 保存为素材 on the task detail, or a batch save by task.
Strong motion, strong sound, a clear face, dense cuts, and good image quality are only parts of the recommendation. They do not replace an MCN team's judgment, and they are not a payment or a view count. Template weights can be adjusted and saved before upload, but raising one weight only changes how that strategy recommends. It does not move the boundary that team assets exist only after review.
The task status can show the run as queued, analyzing, completed, or failed. Completed means the analysis step finished, not that every recommendation was accepted. Failed means that run did not complete. Neither status posts the video anywhere else. The compare page can later show parameter differences and aligned segment differences between two same-source tasks, which is how a team inspects two recommendation passes without treating either pass as already published.
Does a high score put the clip in the library automatically?
No. A recommended clip remains a pending clip until it is accepted. The library requires a separate save as an asset, or a batch save by task.
Do template weights guarantee that a genre will be chosen?
No. Weights only change how the five signals are emphasized. A recommendation can still be rejected, and the weights do not imply views.