The Pipeline Architecture
Sphot's production pipeline is a multi-stage system that transforms raw footage into published Instagram Reels. Here's how each stage works.
Stage 1: Footage Ingestion
When you upload video files to your connected Google Drive folder, Sphot detects the new files automatically. The system downloads the footage, extracts key frames at 2 fps, and prepares them for analysis.
Stage 2: AI Perception
This is where the magic starts. Our perception engine analyzes every extracted frame using multi-modal AI:
- Scene Detection: Identifies distinct scenes and their boundaries
- Object Recognition: Catalogs objects, brands, locations, and text visible in the footage
- Emotion Analysis: Reads facial expressions and body language to understand the emotional arc
- Movement Tracking: Detects action sequences, transitions, and high-energy moments
The output is a structured "footage map" — a complete understanding of what your video contains.
Stage 3: Script Generation
Using the footage map, our content AI writes a complete Reel script:
- Hook: A scroll-stopping opening line (the most important 0.5 seconds)
- Narrative: The core message or story, structured for short-form attention spans
- CTA: A clear call-to-action that drives engagement
Scripts are generated in your preferred language (Hindi, English, or bilingual) and optimized for Instagram's algorithm.
Stage 4: Video Assembly
The ffmpeg pipeline takes over, assembling the final Reel:
- Selects the best clips based on the footage map
- Times cuts to match the script and music tempo
- Applies color grading and exposure correction
- Adds text overlays using Montserrat ExtraBold
- Generates and syncs subtitles
For premium output, the After Effects pipeline via Nexrender applies professional motion graphics templates with dynamic text, transitions, and branded elements.
Stage 5: Music Selection
Our music intelligence system selects a track that matches:
- The emotional tone of the content
- The tempo and energy level
- The target audience and niche
Beat detection ensures cuts align with musical accents for a professional feel.
Stage 6: Quality Control
Every Reel passes through automated QC:
- Resolution and bitrate validation (1080x1920, optimal encoding)
- Audio level normalization (-14 LUFS target)
- Subtitle timing accuracy check
- Brand consistency verification
- Instagram format compliance
Failed items are automatically re-queued with adjusted parameters.
Stage 7: Publishing
The final Reel is published via Instagram's Graph API at the optimal posting time for your audience. Caption, hashtags, and posting schedule are all automated.
The Result
From footage upload to published Reel: under 10 minutes. Compare that to 3-4 hours of manual work.
And the system runs 24/7. Upload footage before bed, wake up to published Reels with real engagement.