AI Social Media Content

Create social media content for all platforms via inference.sh CLI.

Quick Start

Requires inference.sh CLI (belt). Install instructions

belt login

# Generate a TikTok-style video
belt app run google/veo-3-1-fast --input '{
  "prompt": "POV walking through a neon-lit Tokyo street at night, vertical format 9:16, cinematic"
}'

Platform Formats

Platform	Aspect Ratio	Duration	Resolution
TikTok	9:16 vertical	15-60s	1080x1920
Instagram Reels	9:16 vertical	15-90s	1080x1920
Instagram Feed	1:1 or 4:5	-	1080x1080
YouTube Shorts	9:16 vertical	<60s	1080x1920
YouTube Thumbnail	16:9	-	1280x720
Twitter/X	16:9 or 1:1	<140s	1920x1080

Content Workflows

TikTok / Reels Video

# Generate trending-style content
belt app run google/veo-3-1-fast --input '{
  "prompt": "Satisfying slow motion video of paint being mixed, vibrant colors swirling together, vertical 9:16, ASMR aesthetic, viral TikTok style"
}'

Instagram Carousel Images

# Generate cohesive carousel images
for i in 1 2 3 4 5; do
  belt app run falai/flux-dev --input "{
    \"prompt\": \"Minimalist lifestyle flat lay photo $i/5, morning coffee routine, neutral tones, Instagram aesthetic, consistent style\"
  }" > "carousel_$i.json"
done

YouTube Thumbnail

# Eye-catching thumbnail
belt app run falai/flux-dev --input '{
  "prompt": "YouTube thumbnail, shocked face emoji, bright yellow background, bold text area on right, attention-grabbing, high contrast, professional"
}'

Twitter/X Visual Post

# Generate image for tweet
belt app run falai/flux-dev --input '{
  "prompt": "Tech infographic style image showing AI trends, modern design, data visualization aesthetic, shareable"
}'

# Post with Twitter automation
belt app run twitter/post-tweet --input '{
  "text": "The future of AI is here. Here are the top 5 trends reshaping tech in 2024 🧵",
  "media_url": "<image-url>"
}'

Talking Head Content

# 1. Write script with Claude
belt app run openrouter/claude-sonnet-45 --input '{
  "prompt": "Write a 30-second engaging script about productivity tips for a TikTok. Conversational, hook in first 3 seconds."
}' > script.json

# 2. Generate voiceover
belt app run infsh/kokoro-tts --input '{
  "prompt": "<script>",
  "voice": "af_sarah"
}' > voice.json

# 3. Create AI avatar
belt app run bytedance/omnihuman-1-5 --input '{
  "image_url": "https://your-avatar.jpg",
  "audio_url": "<voice-url>"
}'

Content Type Templates

Trending/Viral Style

belt app run google/veo-3 --input '{
  "prompt": "Satisfying compilation style video, oddly satisfying content, smooth transitions, ASMR quality, vertical 9:16"
}'

Tutorial/How-To

belt app run google/veo-3-1 --input '{
  "prompt": "Hands demonstrating a craft tutorial, overhead shot, clean workspace, step-by-step motion, warm lighting, vertical format"
}'

Product Showcase

belt app run bytedance/seedance-1-5-pro --input '{
  "prompt": "Product unboxing aesthetic, sleek packaging reveal, soft lighting, premium feel, satisfying unwrap, vertical 9:16"
}'

Lifestyle/Aesthetic

belt app run google/veo-3-1-fast --input '{
  "prompt": "Day in my life aesthetic, morning routine montage, golden hour lighting, cozy apartment, coffee steam rising, vertical format"
}'

Behind the Scenes

belt app run google/veo-3-1-fast --input '{
  "prompt": "Behind the scenes of creative workspace, artist at work, authentic candid moments, documentary style, vertical 9:16"
}'

Caption & Hashtag Generation

# Generate engaging caption
belt app run openrouter/claude-haiku-45 --input '{
  "prompt": "Write an engaging Instagram caption for a sunset beach photo. Include a hook, value, and call to action. Add 10 relevant hashtags."
}'

Hook Formulas

belt app run openrouter/claude-haiku-45 --input '{
  "prompt": "Generate 5 viral TikTok hooks for a video about morning routines. Use proven patterns like: curiosity gap, bold claim, relatable struggle, before/after, or tutorial format."
}'

Multi-Platform Repurposing

Long to Short Pipeline

# Take a concept and create multiple formats
CONCEPT="productivity hack: 2-minute rule"

# TikTok vertical
belt app run google/veo-3-1-fast --input "{
  \"prompt\": \"$CONCEPT visualization, vertical 9:16, quick cuts, text overlays style\"
}"

# Twitter square
belt app run falai/flux-dev --input "{
  \"prompt\": \"$CONCEPT infographic, square format, minimal design, shareable\"
}"

# YouTube thumbnail
belt app run falai/flux-dev --input "{
  \"prompt\": \"$CONCEPT thumbnail, surprised person, bold text space, 16:9\"
}"

Batch Content Creation

# Generate a week of content
TOPICS=("morning routine" "productivity tips" "coffee aesthetic" "workspace tour" "night routine")

for topic in "${TOPICS[@]}"; do
  belt app run google/veo-3-1-fast --input "{
    \"prompt\": \"$topic content for social media, aesthetic, vertical 9:16, engaging\"
  }" > "content_${topic// /_}.json"
done

Best Practices

1. Hook in first 3 seconds - Start with most engaging moment
2. Vertical first - 9:16 for TikTok, Reels, Shorts
3. Consistent aesthetic - Match brand colors and style
4. Text-safe zones - Leave space for platform UI elements
5. Trending audio - Add popular sounds separately
6. Batch create - Generate multiple pieces at once

Platform-Specific Tips

TikTok

• Fast cuts, trending sounds
• Text overlays important
• Hook immediately

Instagram

• High visual quality
• Carousel for engagement
• Aesthetic consistency

YouTube Shorts

• Clear value proposition
• Subscribe CTAs work
• Repurpose longer content

Twitter/X

• Single striking image
• Controversial hooks work
• Thread potential

Related Skills

# Video generation
npx skills add inference-sh/skills@ai-video-generation

# Image generation
npx skills add inference-sh/skills@ai-image-generation

# Twitter automation
npx skills add inference-sh/skills@twitter-automation

# Text-to-speech for voiceovers
npx skills add inference-sh/skills@text-to-speech

# Full platform skill
npx skills add inference-sh/skills@infsh-cli

Browse all apps: belt app list