Building an AI rap video workflow with scene previews and recoverable jobs

说明与用途

RapVideo 的工作流新增场景预览步骤、原子信用预留以及共享恢复函数，以防止因超时导致的信用损失和视频不完整。

问题

Requests to AI video providers can time out after the provider accepts a job, leaving users unaware of completion and potentially costing credits; generated scenes may have incorrect faces, requiring a review step before animation.

结果

The updated workflow introduces a scene preview step to catch face blending errors, reserves credits atomically with task creation to avoid orphaned jobs, and uses a shared reconciliation function to recover from timeouts and ensure tasks settle correctly.

工作流

• Submit portraits and settings
• Generate scene image
• User reviews scene image
• Approve or reject scene
• Generate video with audio using approved scene
• Store MP4 and task history
• Reconcile pending tasks via polling, webhook, or scheduled recovery

证据

A request can time out after the provider has already accepted the job. The video might then finish after the user closes the tab.

来源

The user goes through this sequence: portrait(s) + scene choices ↓ scene image generation ↓ user checks the scene ↓ video generation with audio ↓ stored MP4 + task history

来源

Before submission, the application reserves the task and debits credits in one atomic database operation.

来源