Who This Is For
This feature is designed for:
Automation engineers and operations managers who need to extract data, summaries, or verification checks from video content without manual review.
Marketing and fulfillment teams requiring automated content moderation or verification of video assets generated in prior workflow steps.
Support teams automating the ingestion of customer video submissions from forms or webhooks.
What is the AI Analyze Video Action?
The AI analyze video action is an automated processing step in the Workflow & Automation System that accepts a video URL and natural language prompt, evaluates both visual and audio data, and outputs a structured text response.
Use this action when an automation requires automated transcription, timestamp extraction, visual confirmation, or content summarization from video files. Do not use this action to edit, render, or host video files directly.
Which Video Formats and URLs Are Supported?
The action processes publicly accessible video links and direct media files. Each action handles one video URL per execution.
Supported input formats include:
Public YouTube URLs, including standard watch links, shortened youtu.be links, and YouTube Shorts.
Direct public links to video file formats: MP4, MOV, WebM, MPEG, FLV, WMV, and 3GP.
Dynamic URL variables mapped from incoming webhooks, form submissions, or the AI Video Generation action.
Links from platforms such as Vimeo and TikTok are not supported because those platforms block automated crawler access.
Which AI Models Are Available?
The action supports two Google Gemini models with different processing capabilities and speeds:
Gemini 3.8 Flash: The default model, optimized for detailed multimodal analysis and complex prompts.
Gemini 3.5 Flash Lite: A lightweight model designed for faster turnaround on straightforward extraction and classification tasks.
How Does Pricing and Token Usage Work?
The AI analyze video action is a premium workflow action billed per execution based on total token consumption.
Execution costs vary depending on three operational variables:
Model: The specific model selected in the action configuration.
Video length: Longer durations increase video and audio token consumption.
Response length: Long text generations consume more output tokens than short, single-word, or boolean answers.
How to Configure the AI Analyze Video Action
Follow these steps to insert and configure the action inside a workflow:
Navigate to Automation > Workflows in your account.
Open an existing workflow canvas or create a new workflow.
Click the add step icon (+) and select AI analyze video from the AI Actions group.
Enter an identification label in the ACTION NAME field.
Select either Gemini 3.8 Flash or Gemini 3.5 Flash Lite from the MODEL dropdown menu.
Provide a valid URL in the VIDEO field by typing a static link or clicking the tag icon to select a dynamic variable from a prior step.
Enter instructions in the ANALYSIS PROMPT field defining the target information and required output structure.
Click Save action to retain the configuration.
How to Test the Video Analysis Action
The workflow builder provides a test tab to validate video ingestion and prompt accuracy before deployment.
Open the configured AI analyze video panel.
Click the Test tab located next to Setup.
Click Test Action to execute the query against the provided URL.
Review the generated text inside the Response window to confirm accuracy and structure.
Confirm that the downstream workflow steps correctly reference the Response output key.
Common Issues and Frequently Asked Questions
Why does the action fail when using a Vimeo or TikTok link?
Vimeo and TikTok block external automated extraction requests. You must provide a direct public video file link (such as an MP4 hosted in Media Storage) or a public YouTube link.
Why is the action taking long to complete?
Multimodal processing time scales with the length of the video. While a short clip processes in a few seconds, a 10-minute video typically requires approximately 30 seconds to parse audio and visual data.
Why does the generated text contain unwanted Markdown syntax?
Google Gemini outputs Markdown formatting by default. If your downstream steps require plain text for SMS or database fields, explicitly state "Return plain text only without markdown formatting, headers, or bullet characters" in the ANALYSIS PROMPT field.
Can one action step analyze multiple videos at the same time?
No. Each execution of the AI analyze video action is limited to a single video URL. To analyze multiple files, configure sequential AI analyze video steps or iterate through an array using workflow branching.