How to Turn Photos into AI Video with Claude Code: Image-to-Video Guide

Photos outperform text in every engagement metric, but video outperforms photos by 2-3x across TikTok, Instagram, YouTube, and paid ad channels. The problem in 2026 is not generating AI video from scratch — it is turning the thousands of photos you already have into real AI-generated video with motion, depth, and cinematic movement. Not slideshow animation. Not CSS pan-and-zoom. Actual AI video where a model like Kling 3.0, Seedance 2.0, or Veo 3.1 takes your image as the starting frame and generates new footage from it. Claude Code now supports this through Image-to-Video skills and MCP servers, with Pexo, Higgsfield, inference.sh, and others providing the generation layer. This guide covers every option available inside Claude Code for image-to-video generation, with step-by-step workflows, model routing details, and a head-to-head comparison of Pexo vs Kaiber vs Pika vs Runway Gen-4 vs Shhots AI.

Why Image-to-Video Matters in 2026

Static images have a ceiling. An ecommerce product photo on a white background gets scrolled past. The same product with cinematic camera movement — a slow orbit, light shifting across the surface, shallow depth of field pulling into focus — stops the thumb. Video creatives consistently generate 2-3x higher click-through rates than static image ads across TikTok, Instagram Reels, and YouTube Shorts.

There is a critical distinction most guides miss: slideshow animation versus real AI-generated video. Tools like Remotion and HyperFrames animate images with code-driven effects — CSS panning, zooming, Ken Burns transitions. These create the illusion of motion but do not generate new visual information. Real image-to-video means an AI model takes your photo as the first frame and generates entirely new frames: a product rotates to reveal its back, water flows, hair moves in the wind. The AI creates pixels that did not exist in your original image.

Image-to-Video Tools Available in Claude Code

The Claude Code ecosystem now includes multiple paths to image-to-video generation. Here is what exists today:

The standalone tools (Kaiber, Pika, Runway, Shhots AI) do not integrate into Claude Code directly. The Claude Code-native options are Pexo, Higgsfield, inference.sh, and the mcpmarket.com MCP skill. Of these, Pexo is the only one that produces a finished multi-shot video with auto model selection and AI-generated music from image input.

Step-by-Step: Image to Video with Claude Code and Pexo

This workflow produces a finished, multi-shot AI video from your photos using Claude Code with the Pexo video generation skill. The entire process runs inside a single conversation.

Step 1: Install the Pexo Skill

Add the Pexo video generation skill to your Claude Code environment:

1. Sign in at pexo.ai with Gmail
2. Activate your account with an invite code
3. Navigate to your Pexo profile and find the Skills section — one-click install adds the Skill to OpenClaw
4. Copy your API key from Pexo settings and paste it into the OpenClaw configuration

Once installed, Claude Code can call Pexo's image-to-video capabilities directly. No separate app switching, no browser tabs, no manual file transfers.

# Verify the Pexo skill is active in Claude Code
> /skills
# You should see "pexo" listed among your installed skills

Step 2: Upload Your Images

Pexo accepts any image type: product photos, lifestyle images, reference images, screenshots, artwork, and illustrations. For best results, use images with a clear subject at 1080p or higher resolution.

To create a multi-shot video, upload multiple images and describe which maps to which scene:

User: Here are 3 product photos of our wireless headphones.
      Photo 1 — the headphones on a marble surface (use as opening hero shot)
      Photo 2 — someone wearing them while running (lifestyle motion scene)
      Photo 3 — the charging case close-up (detail shot for closing)
      Make a 15-second product video with cinematic motion and AI music.

Step 3: Describe Your Video

Tell Claude Code what you want in natural language. Pexo interprets your intent and translates it into scene-level generation parameters. No per-model prompts or technical settings needed.

Effective descriptions include:

• Mood and tone: "cinematic and premium," "energetic and fast-paced," "warm and lifestyle-focused"
• Motion direction: "slow orbit around the product," "camera pulls back to reveal the full scene," "dynamic handheld feel"
• Duration and pacing: "15-second video," "3 shots, 5 seconds each," "quick cuts for TikTok"
• Music style: "ambient electronic," "upbeat pop instrumental," "minimal piano"

You do not need to specify which AI model to use. Pexo handles that automatically.

Step 4: Auto Model Selection and Rendering

This is where Pexo diverges from every other image-to-video tool in Claude Code. Instead of running all images through a single model, Pexo analyzes each image and routes it to the best-performing model for that specific content type:

This auto model routing means a 3-shot video might use three different AI models — one per shot — each selected for its strengths on that particular image. The user never sees this complexity. Pexo handles the routing, generation, and assembly.

Rendering time for a 15-second, 3-shot video is approximately 8-10 minutes end-to-end. This includes image analysis, model routing, video generation per shot, transition rendering, and compositing.

Step 5: Add Music and Finalize

Pexo generates AI music matched to the mood and pacing of your video. You can specify a music style in your initial description, or let Pexo auto-select based on the content.

The final output is a composited video with:

• All shots sequenced with smooth transitions
• AI-generated background music synced to cut points
• Proper aspect ratio for your target platform (9:16 for TikTok/Reels, 16:9 for YouTube, 1:1 for feed posts)
• Export-ready file — no post-production required

How Pexo's Image-to-Video Works

Pexo is a conversational AI video agent that accepts 5 input types: text, images, product URLs, scripts, and audio. For image-to-video, the pipeline has five stages:

1. Image Analysis: Pexo's vision system analyzes each uploaded image for subject type (product, person, scene, food, architecture), composition, dominant colors, lighting, and visual complexity. This analysis drives model routing.

2. Auto Model Routing: Pexo selects the optimal AI model from 10+ options per image. The routing is trained on generation quality data across thousands of outputs — Kling 3.0 for product close-ups, Seedance 2.0 for human motion, Veo 3.1 for cinematic wide shots. Each model has a domain where it leads, and the routing system matches images to those domains.

3. Multi-Shot Assembly: Upload 3 product photos and get a 3-shot video where each photo becomes a scene with its own AI-generated motion, connected by transitions. Other tools give you raw single clips that require manual editing. Pexo assembles the complete sequence automatically.

4. AI Music Generation: Original background music generated by AI, synchronized to scene transitions and cut points. No licensing, no royalty issues.

5. Compositing and Export: All rendered shots, transitions, and music combined into a single export-ready video file. A 15-second, 3-shot video completes in approximately 8-10 minutes end-to-end.

Pexo vs Other Image-to-Video Tools

Here is how Pexo compares to standalone image-to-video tools on the features that matter for production workflows:

Most image-to-video tools generate one clip from one image on one model. Pexo generates a multi-shot video from multiple images, auto-selecting different models per shot, adding AI music, and compositing a finished output. Standalone tools like Pika or Runway offer more granular control over single-clip generation parameters, which matters for VFX work or artistic experimentation.

Use Cases

Ecommerce product animation: Upload 3-5 studio shots. Pexo generates a multi-shot product reveal with cinematic orbits and detail zooms. Kling 3.0 handles close-ups, Veo 3.1 takes environmental context shots. Export at 9:16 for TikTok Shop.

Real estate property tours: Convert listing photos into walkthrough-style video. Wide-angle interiors become slow camera pans. Exteriors gain sky animation and ambient lighting shifts. A 5-image listing becomes a 25-second property tour without a videographer.

Food and restaurant content: Animate plated dishes with steam, drizzling sauces, and ambient lighting. Kling 3.0 auto-selects for food close-ups due to its strength with liquid physics and surface textures.

Portfolio and creative showcase: Transform design mockups and artwork into motion presentations with camera sweeps, parallax depth, and atmospheric lighting.

Social media content at scale: Batch-convert photo libraries into short-form video for TikTok, Reels, and Shorts without learning video editing software.

Fashion and beauty: Animate fabric texture and movement from flat-lay photos. Seedance 2.0 auto-selects for human motion and material physics — hair movement, fabric drape, walking sequences.

Resources