Automatically turn podcasts to shorts

What This Agent Does

This powerful automation workflow transforms raw video content into viral-ready TikTok posts with AI-generated captions and titles—all without manual editing. Simply submit your main video clip and background footage, and the agent automatically transcribes your content, generates perfectly-timed captions, assembles a professional video timeline, renders the final product, and publishes directly to TikTok.

Save 2-3 hours per video by eliminating manual caption timing, video editing, and the tedious upload process. This workflow is perfect for content creators, social media managers, and marketing teams who need to maintain a consistent TikTok presence without spending hours in editing software. Whether you're repurposing long-form content, creating educational snippets, or producing entertainment clips, this agent handles the entire production pipeline from raw footage to published post.

Key benefits include: automated caption generation with precise timing, professional video composition with background overlays, AI-powered viral title creation, and seamless TikTok publishing—all executed in a single automated sequence.

Required Integrations

Shotstack

Shotstack is a cloud-based video editing API that powers the video processing capabilities of this workflow. You'll need it to ingest video sources, build the timeline with captions and overlays, and render the final TikTok-ready video.

Setup steps:

1. Visit shotstack.io and create a free account
2. Navigate to your dashboard and locate the API Keys section
3. Copy your API Key (it will start with a long alphanumeric string)
4. Note whether you're using the sandbox or production environment (sandbox is free for testing)
5. In TaskAGI, go to Integrations → Add Integration → Shotstack
6. Paste your API key into the authentication field
7. Select the appropriate environment (sandbox for testing, production for live videos)
8. Click Connect to verify the integration

Important: Sandbox renders include a watermark and have usage limits. Upgrade to production when you're ready to publish watermark-free videos.

UploadPost

UploadPost enables direct publishing to TikTok and other social media platforms. This integration eliminates the need to manually download and upload your rendered videos.

Setup steps:

1. Go to uploadpost.io and sign up for an account
2. Connect your TikTok account by clicking Add Social Account → TikTok
3. Follow the OAuth flow to authorize UploadPost to post on your behalf
4. Once connected, navigate to API Settings in your UploadPost dashboard
5. Generate a new API Key and copy it securely
6. In TaskAGI, go to Integrations → Add Integration → UploadPost
7. Paste your API key and click Authenticate
8. Select which TikTok account to use if you have multiple connected

Note: Ensure your TikTok account has creator permissions and meets TikTok's API access requirements. Business accounts typically have better API support.

Configuration Steps

Step 1: Video Input Form (Trigger)

This form node collects the necessary inputs to start your workflow. Configure the following fields:

• Main Video URL: A text input field where users paste the URL of the primary video clip (the main content)
• Background Video URL: A text input field for the background video that will play behind the main clip
• Video Description: A text area for context about the video content (helps the AI generate better captions and titles)

Example configuration:

Field 1: main_video_url (type: text, required: true)
Field 2: background_video_url (type: text, required: true)
Field 3: description (type: textarea, required: false)

The form submission timestamp automatically flows to the next node to initiate processing.

Step 2: Ingest Main Clip

This Shotstack node downloads and prepares your main video for editing.

• Source URL: Map to {{trigger.main_video_url}} from the form
• Source Type: Set to video
• Wait for Completion: Enable this option (checked) to ensure the video is fully ingested before proceeding

The node outputs a source_id that uniquely identifies your ingested video in Shotstack's system.

Step 3: Ingest Background Video

Configure identically to the main clip ingestion:

• Source URL: Map to {{trigger.background_video_url}}
• Source Type: video
• Wait for Completion: Enabled

This creates a second source_id for the background footage.

Step 4: Generate Timed Captions

This AI Agent node uses GPT-4o to transcribe your video and create precisely-timed captions.

Model: gpt-4o (required for accurate timing and transcription)

System Prompt:

You are a video caption specialist. Analyze the provided video content and generate timed captions in JSON format. Each caption should include start time, end time, and text. Format captions for maximum readability on mobile devices (max 15 words per caption). Return ONLY valid JSON.

User Prompt:

Generate timed captions for this video: {{trigger.description}}

Return format:
{
  "captions": [
    {"start": 0.0, "end": 2.5, "text": "Caption text here"},
    {"start": 2.5, "end": 5.0, "text": "Next caption"}
  ]
}

Input Data: Map the main video source_id: {{ingest_main_clip.source_id}}

The AI will analyze the video and output structured caption data with precise timing.

Step 5: Build Shotstack Timeline

This Function node constructs the video editing timeline using JavaScript.

Function Code:

const captions = JSON.parse(input.captions);
const mainClip = input.main_source_id;
const bgClip = input.background_source_id;

const timeline = {
  tracks: [
    {
      clips: captions.captions.map(cap => ({
        asset: {
          type: "html",
          html: `<p>${cap.text}</p>`,
          css: "p { color: white; font-size: 48px; font-weight: bold; text-align: center; text-shadow: 2px 2px 4px black; }"
        },
        start: cap.start,
        length: cap.end - cap.start,
        position: "bottom"
      }))
    },
    {
      clips: [{
        asset: { type: "video", src: mainClip },
        start: 0,
        length: "auto"
      }]
    },
    {
      clips: [{
        asset: { type: "video", src: bgClip },
        start: 0,
        length: "auto",
        fit: "cover",
        opacity: 0.3
      }]
    }
  ],
  output: {
    format: "mp4",
    resolution: "sd",
    aspectRatio: "9:16"
  }
};

return { timeline: timeline };

Input Mapping:

• captions: {{generate_captions.result}}
• main_source_id: {{ingest_main_clip.source_id}}
• background_source_id: {{ingest_background.source_id}}

This creates a three-track timeline: captions on top, main video in the middle, and dimmed background video at the bottom.

Step 6: Render TikTok Video

Configure the Shotstack render node:

• Timeline: Map to {{build_timeline.timeline}}
• Output Format: mp4
• Resolution: sd (standard definition is sufficient for TikTok)
• Aspect Ratio: 9:16 (vertical video for TikTok)

The node returns a render_id for tracking the render progress.

Step 7: Wait for Render Complete

This polling node checks render status:

• Render ID: {{render_video.render_id}}
• Poll Interval: 5 seconds
• Max Attempts: 60 (allows up to 5 minutes for rendering)

The node waits until the status is done before proceeding, then outputs the final video URL.

Step 8: Generate Viral Title

Another AI Agent node creates an engaging TikTok title:

Model: gpt-4o

System Prompt:

You are a viral TikTok content strategist. Create compelling, attention-grabbing titles that drive engagement. Use hooks, curiosity gaps, and trending language. Keep titles under 100 characters.

User Prompt:

Create a viral TikTok title for this video: {{trigger.description}}

Include relevant hashtags and emojis. Make it scroll-stopping.

The AI generates an optimized title designed to maximize views and engagement.

Step 9: Post to TikTok

Final node publishes your video:

• Video URL: {{wait_for_render.url}}
• Title: {{generate_title.result}}
• Privacy: Choose public, friends, or private
• Allow Comments: true (recommended for engagement)
• Allow Duet: true (increases viral potential)
• Allow Stitch: true

The workflow completes when the video is successfully posted to TikTok.

Testing Your Agent

Running Your First Test

1. Use test videos: Start with short clips (15-30 seconds) to minimize processing time
2. Provide clear descriptions: The more context you give, the better the AI-generated captions and titles
3. Use Shotstack sandbox: Enable sandbox mode to avoid production costs during testing
4. Monitor execution: Watch the workflow progress in real-time through the TaskAGI execution panel

Verification Checklist

After form submission:

• ✓ Verify both video URLs are accessible and valid
• ✓ Check that the workflow status changes to "Running"

During video ingestion:

• ✓ Confirm both source_id values are generated (long alphanumeric strings)
• ✓ Check for any ingestion errors (invalid URLs, unsupported formats)

After caption generation:

• ✓ Review the JSON output to ensure captions are properly formatted
• ✓ Verify timing values are logical (end time > start time)
• ✓ Check that caption text is readable and accurate

During rendering:

• ✓ Monitor the render status (queued → rendering → done)
• ✓ Typical render time: 1-3 minutes for a 30-second clip
• ✓ Verify the output URL is accessible

Before TikTok posting:

• ✓ Preview the generated title for quality and appropriateness
• ✓ Ensure the video URL is a direct link to the MP4 file
• ✓ Confirm your TikTok account is properly connected

Success indicators:

• All nodes show green checkmarks
• Final node returns a TikTok post ID or URL
• Video appears in your TikTok drafts or published feed

Troubleshooting

Video Ingestion Failures

Error: "Unable to ingest source - invalid URL"

• Solution: Ensure video URLs are direct links to video files, not webpage links. Use services like Dropbox, Google Drive (with direct download links), or S3 buckets.

Error: "Source format not supported"

• Solution: Shotstack supports MP4, MOV, and WebM. Convert videos to MP4 for best compatibility.

Caption Generation Issues

Error: "AI response is not valid JSON"

• Solution: The AI occasionally returns explanatory text. Update the system prompt to emphasize "Return ONLY valid JSON with no additional text or markdown formatting."

Problem: Captions are poorly timed or inaccurate

• Solution: Provide more detailed video descriptions. Consider using a video with clearer audio. GPT-4o performs better with high-quality audio sources.

Rendering Problems

Error: "Render failed - timeline validation error"

• Solution: Check the Build Timeline function output. Ensure all asset URLs are valid and the JSON structure matches Shotstack's schema.

Problem: Render stuck in "queued" status

• Solution: Shotstack sandbox has limited concurrent renders. Wait a few minutes or upgrade to production. Check Shotstack's status page for service issues.

TikTok Publishing Failures

Error: "Authentication failed"

• Solution: Reconnect your TikTok account in UploadPost. TikTok tokens expire periodically and need refresh.

Error: "Video does not meet TikTok requirements"

• Solution: Ensure video is 9:16 aspect ratio, between 3 seconds and 10 minutes, and under 287MB. Check TikTok's current upload specifications.

Problem: Video posts but doesn't appear

• Solution: Check if the video was posted as a draft. Some TikTok accounts require manual approval before publishing. Verify privacy settings in the Post to TikTok node.

Next Steps

After Successful Setup

Immediate actions:

1. Create a library of background videos for variety in your content
2. Document your video description format for consistency
3. Set up a content calendar to schedule regular workflow executions
4. Monitor your first few posts to refine caption style and title generation

Optimization Suggestions

Improve caption quality:

• Fine-tune the caption generation prompt based on your niche (educational, entertainment, promotional)
• Adjust maximum words per caption for your audience's reading speed
• Experiment with caption positioning and styling in the timeline function

Enhance video production:

• Add brand logos or watermarks by including an image asset in the timeline
• Incorporate intro/outro clips for consistent branding
• Experiment with background video opacity for optimal readability

Boost engagement:

• A/B test different title generation prompts to find what resonates
• Analyze which hashtag combinations perform best
• Create variations of the workflow for different content types

Advanced Usage Tips

Batch processing: Modify the trigger to accept multiple video URLs and use a loop node to process several videos in sequence.

Content repurposing: Connect this workflow to a YouTube download node to automatically convert long-form content into TikTok clips.

Analytics integration: Add a final node that logs video performance data to a spreadsheet or database for tracking ROI.

Multi-platform publishing: Duplicate the UploadPost node to simultaneously publish to Instagram Reels, YouTube Shorts, and other platforms.

Dynamic backgrounds: Create a library of background videos tagged by category, then use AI to select the most appropriate background based on video content.

Custom branding: Modify the timeline function to include your brand colors, fonts, and logo positioning for consistent visual identity across all posts.

This workflow represents a complete video production pipeline that scales with your content needs. Start with the basic configuration, then gradually add sophistication as you identify opportunities for improvement specific to your audience and content strategy.