Automate the end-to-end creation of UGC-style video ads—from script intake to final stitched delivery.
It ingests the campaign script and character/setting descriptions from Google Sheets and converts them into a full production plan. It generates assets, scripts scenes with consistent visuals, renders clips, and stitches them into a final UGC video. It uploads the finished video to Drive and updates Sheets with per-scene and final delivery links.
The AI agent performs concrete actions to produce scalable ads.
Ingests inputs from Google Sheets.
Generates selfies and assets via Kie API.
Analyzes visuals for consistency with Claude-4o-mini and Claude Opus.
Converts scripts into structured scenes with environment-aware prompts.
Renders video clips for each scene with Kie VEO3 and monitors completion.
Stitches clips into a final video with fal.ai FFmpeg and updates Sheets.
Two sentences of explanation.
Three-step system flow in plain language.
Collects inputs from Google Sheets and defines the asset, scene, and rendering plan.
Creates selfie prompts, outputs visuals via Kie, analyzes for continuity with Claude tools, and formats 8-second scenes.
Renders clips, stitches into a final video with fal.ai, uploads to Drive, and updates Sheets with final links.
One realistic scenario.
Scenario: A cosmetics brand wants a 60-second UGC-style ad. Script and character notes are stored in Google Sheets. The AI agent generates 6 scenes of about 10 seconds each, creates corresponding visuals, renders video clips via Kie, stitches them into a single 60-second ad with fal.ai FFmpeg, uploads the final video to Drive, and writes the final links back to Sheets. Estimated time: about 90 minutes.
One supporting sentence.
Need scalable UGC ads with consistent visuals and fast turnaround.
Secure automated tooling to generate assets and scripts without manual bottlenecks.
Automate repetitive tasks to handle multiple clients efficiently.
Gain per-scene visibility and control over delivery timelines.
Produce high-quality ads quickly without large studios.
Maintain creative guardrails while scaling output.
One supporting sentence with short explanation.
Generates selfie visuals and video clips (Nano Banana and VEO3) inside the AI agent.
Performs prompt engineering and scene scripting for production flow.
Analyzes images for visual continuity and consistency.
Handles FFmpeg-based stitching of scene clips into the final video.
Stores assets and shares final delivery links.
Controls input, runs status tracking, and captures final links.
One supporting sentence with short explanation.
One supporting sentence with short explanation.
You need a script and character/setting description in your Google Sheet, plus an aspect ratio and target video length. The AI agent reads these inputs, validates them, and creates a scene plan. It then fetches assets and prompts from integrated tools to begin production. You can also specify regional variants or prompts to guide creative direction. If inputs are missing, the agent reports the gaps in Sheets for quick resolution.
Typical turnaround for a single 60-second video with 5–6 scenes is around 60–120 minutes, depending on asset complexity and network latency with external APIs. The agent processes scenes in parallel where possible and streams results back to Sheets as they complete. You can monitor progress in the input sheet and via per-scene status updates. For larger campaigns, batch runs can be scheduled to run overnight. Final delivery links appear in Sheets once stitching completes.
Yes. Scene lengths and transitions are defined in the production plan and can be adjusted in Sheets before a run. The AI agent respects those parameters when generating prompts and sequencing clips. You can re-run individual scenes if needed, updating their status to Redo. Changes propagate through rendering, stitching, and final delivery as part of the same AI agent workflow.
The agent outputs standard MP4 video files at common resolutions (e.g., 1080p). You can specify aspect ratio and resolution in the input sheet. The final video is stored in Google Drive with shareable links and can be re-encoded for other platforms if required. Audio tracks are kept in sync with video, and keyframe pacing is preserved across scenes. Additional variants can be produced by duplicating the run with adjusted prompts.
Quality is ensured through multiple checks: image analysis with Claude-4o-mini for visual continuity, environment-aware prompts from Claude Opus, and automated validation of scene timing. Individual scenes can be retried by setting LAUNCH = Redo in the per-scene sheet, ensuring only failed parts are reprocessed. The stitching step verifies that all required scenes are present before final delivery. Manual review can be requested for final approval if needed.
Yes. Failed scenes can be retried by marking the corresponding row as Redo in the Video Data sheet. The AI agent reprocesses only those scenes, leaving successful ones intact. This avoids re-running the entire workflow and speeds up iterative edits. You receive updated links once stitching completes for the re-run results.
Progress is tracked in Google Sheets with statuses at both the campaign level and per-scene. The AI agent updates these statuses automatically as scenes are created, rendered, and stitched. You can trigger workflow orchestration from other workflows if needed. Notifications can be configured to alert your team when milestones are reached or if failures occur.
Automate the end-to-end creation of UGC-style video ads—from script intake to final stitched delivery.
