No
Node-based AI video editing workbenchAI video editing workflow platform · V62
Connecting to the live product service
AI video editing workflow platform

Split videos into editable, observable, and reusable AI editing workflows

After the user uploads the video, he configures and connects professional AI nodes in the canvas, runs a traceable editing workflow, and produces reviewable attention curves, short video candidate clips, multilingual homophonic dubbing solutions, and material/editing operations that can be inserted into the timeline.

  1. 01
    Upload videoCreate a video_input record and generate an input node that can be referenced by the canvas.
  2. 02
    orchestration nodeCreate professional AI nodes in canvas and connect them into workflows.
  3. 03
    Run and observeExecute workflows and record the real-time or near-real-time status of each node.
  4. 04
    review resultsReview attention curves, short clip candidates, dubbing plans, and timeline materials.
  5. 05
    Manual approvalAuthorize and confirm compliance for high-risk content such as dubbing with same timbre.
Workflow statusWorkflow has not been run yetCount the number of each status in the current project workflow_graph.run_status and node_run_log.
Number of low attention intervalsNo attention curve yetCount the number of latest attention_curve.low_attention_ranges.
Candidate short fragmentNo candidate segments yetCount the number of clip_candidate records in the current project.
Dubbing plan pending approvalNo items to be approvedCount the number of voice_dub_plans with approval_status=pending_human_approval.
BUSINESS INPUT

New project

Each project represents a video editing project to be processed, including source video, target language, short clip retrieval intention, voiceprint authorization confirmation and workflow preferences.

Loading product-specific business fields.

Not saved
TRACEABLE OUTPUT

Workflow results

Results, sources, failures, approvals, and rollbacks are saved by project.

0 live runs
∅

No business results yet

Please upload the source video first and fill in the clipping target before the system can generate an executable node canvas.

PRIVACY

Data use boundary

The videos, audios, voiceprint authorization certificates, translation dictionaries and safe reserved words you upload may contain personal voices, portraits, commercial scripts or unpublished content. This product only uses these data for video analysis, workflow execution, dubbing plan generation, material drafts and audit records of the current project, and will not display them as public cases. Current 10 role review supplements: Classify safewords, translation dictionaries, prompts, transcripts/OCR, generated previews, node input summaries, error logs, approval reasons, source videos, and voice authorization files as sensitive or quasi-sensitive. Logs, exports, audit records, and error messages should store references, summaries, timestamps, and object IDs rather than full videos, full audio, full transcripts, full prompts, full authorization files, credentials, or unnecessary personal data. If OpenAI Responses API or any external AI processor receives project data, disclose which data categories may be sent and mark unknown processor retention/training/regional details as unknown unless proven by implementation or policy. All sensitive records, previews, exports, approvals, and rollback snapshots must be scoped to the current site/workspace; access by bare record ID must not be sufficient.

RETENTION

Data retention

Project files and generation records are retained for 30 days by default; users can delete them in advance. After deletion, the site's workflow records, reports and draft materials will be marked as deleted and entered into the background cleanup queue; for compliance auditing purposes, manual approval logs can be retained for up to 180 days.

SAFETY

Human responsibility and rollback

The attention curves, short clip suggestions, translated dubbing and material drafts generated by this product are all AI-assisted results and may involve recognition errors, translation deviations, unstable generation quality, copyright or voiceprint authorization risks. Users must conduct manual review before use, publication or delivery, especially content with homophonic dubbing and character voices.

Rolling back only means revoking the workflow running results, report status, timeline draft, material recording or approval status in this site to the previous version; this product is not connected to an external production editing system or publishing platform, so it will not and cannot roll back videos, published content or files in external systems.