Social Media Pipeline

Social Media Content Pipeline

Build a creative brand identity and content generation pipeline for social media accounts — from naming through visual identity, content pillars, and automated production workflows.

When to Use

• User wants to start a new social media account (any platform)
• User asks to brainstorm brand names, visual identity, or content strategy
• User wants to set up a content generation pipeline (text + images + video)
• User mentions "social media strategy", "brand identity", "content calendar", "posting schedule"
• User wants to combine AI generation tools (ComfyUI, image_gen) with a publishing workflow

When NOT to Use

• Single image generation request — use comfyui or image_gen directly
• Pure design work (landing page, UI) — use claude-design or sketch
• Technical ComfyUI setup — use comfyui skill
• One-off content idea brainstorming — use ideation

Pipeline Phases

Phase 1: Brand Foundation

Establish identity before generating anything.

1a. Name Discovery

• Use ideation skill's constraint method for name brainstorming
• Consider: pronounceability, cross-platform handle availability, visual brandability
• Test names against: Does it work as a hashtag? As a handle? On a business card?
• Present 6-10 options across 2-3 vibe categories (evocative, bold, technical, etc.)
• Narrow to top 3 with rationale

1b. Visual Identity

• Color palette: 2-3 primary colors + neutrals, with hex codes
• Typography: 1-2 font families (display + body)
• Style direction: the aesthetic "feel" (e.g., "cinematic anime", "clean minimal", "dark moody")
• Mood references: 3-5 existing accounts or artworks that capture the vibe
• Use claude-design skill to prototype visual concepts if needed

1c. Voice & Tone

• How the account talks to its audience
• Content tone: educational, inspirational, technical, playful, etc.
• Engagement style: responsive, curated, mysterious, community-driven

1d. Content Pillars

• 3-5 recurring content types (e.g., "scene of the day", "process breakdown", "before/after")
• Format mix: static images, carousels, threads, video, reels
• Platform-specific adaptations (X vs Instagram vs TikTok)

Phase 2: Content Strategy

2a. Posting Cadence

• Start conservative: 3-5 posts/week minimum viable
• Batch generation beats daily grinding
• Best times to post (platform-specific, but general: evenings, weekends)
• Content calendar: theme days, series, events

2b. Content Types by Platform

2c. Engagement Strategy

• Reply strategy: how fast, how deep
• Community building: retweets, collaborations, features
• Hashtag strategy: niche + broad mix

Phase 3: Technical Pipeline

3a. Generation Setup

• Confirm backend choice: local vs cloud vs mixed
• Mixed/local-cloud examples:
  • Local Illum for speed/iteration on 6–12 GB cards
  • Cloud Flux for consistency, hi-res, and heavier stacks on 48 GB
• Hardware constraints → model selection (see comfyui skill)
• Workflow templates for each content type, one per backend/model

3b. Prompt & Model Routing

• Same brand should not sound identical across different backends; map prompt language to model behavior:
  • Local Illumina: painterly texture words, looser brushwork phrasing, lower CFG
  • Cloud Flux: screenplay-style structure, explicit composition, lower CFG
• Maintain a prompt DNA template so every prompt includes archetype + scene + light + texture + quality anchors
• Save model-specific positive/negative token banks in brand files

3c. Consistency Stack To keep a gallery feeling like one film:

• 1–2 LoRAs at 0.4–0.7 weight
• IP-Adapter FaceID/Plus when character identity must hold across poses
• ControlNet OpenPose/DWPose for pose lock; ControlNet Depth for environment fidelity
• Hires fix with stable denoise strength
• Fixed seed within a character arc; reroll per new scene

3d. Output Organization

outputs/
  raw/          # untouched generations, highest res
  curated/      # selected finals, hero/supporting/environment
  x_ready/      # cropped, branded, compressed, platform crops
prompt_archive/

3e. Blender / 3D Handoff (if brand work will later become 3D)

• Generate with 3D in mind: centered subject, one key light + one fill, neutral camera angle, clear silhouette read, distinct material groupings
• For concept packages: front/side/3-quarter views, palette PNG, materials.txt, pose reference folder
• Use ControlNet depth/lineart on base renders to keep composition fidelity across 2D→3D handoff

Phase 4: Publish & Iterate

4a. Launch

• Profile setup: bio, header, profile pic (use brand identity)
• First 9-12 posts: establish visual consistency
• Pin best work

4b. Feedback Loop

• Track engagement (likes, retweets, replies, follows)
• Identify top-performing content types
• Iterate on style, timing, format

4c. Growth

• Collaborations with other AI artists
• Participate in challenges/trends
• Cross-platform posting
• Build email list or Discord community

Brand Identity Template

## Brand: [Name]

### Visual Identity
- Primary colors: [#hex1, #hex2]
- Accent: #hex3
- Neutral: #hex4
- Font (display): [Font Name]
- Font (body): [Font Name]
- Style keywords: [3-5 adjectives]

### Voice
- Tone: [educational/inspirational/technical/playful]
- Engagement: [responsive/curated/mysterious/community]
- Sample post: [example of typical content]

### Content Pillars
1. [Pillar 1] — [description, frequency]
2. [Pillar 2] — [description, frequency]
3. [Pillar 3] — [description, frequency]

### Platform Handles
- X: @handle
- Instagram: @handle
- TikTok: @handle

Pitfalls

1. Don't generate before branding — inconsistent output wastes content and confuses audience
2. Don't over-post at launch — 3 quality posts/week beats 7 mediocre ones
3. Don't ignore platform differences — X and Instagram need different content
4. Don't skip the feedback loop — post → measure → iterate, not post → post → post
5. Don't lock brand too early — first 2 weeks are exploration, not commitment
6. GPU VRAM matters — verify hardware specs before making claims (I got this wrong once)

Reference Files

• references/brand-brainstorming.md — name discovery workflow, constraint method, example output
• references/content-calendar.md — weekly theme structure, pillar definitions, posting schedule, metrics
• references/loratlas-cheonma.md — Loratlas discovery notes, curated Cheonma-adjacent style tiers, prompt protocol, brand file pointer
• references/brand-markdown-patch-pitfall.md — when patching prompt template markdown, always re-read after patching; small edits can bisect fenced code blocks or duplicate headings

Integration Points

• ideation — name and concept brainstorming
• comfyui — local image/video generation
• image_gen — quick generation without local setup
• claude-design — visual identity prototyping
• sketch — rapid concept exploration
• nous-branding — if the brand is Nous-adjacent
• cronjob — scheduled content generation

Example: AI Art Account Setup

User request: "I want to start an AI art page on X with anime/semi-realistic style"

Execution:

1. Brainstorm 10 names across 3 vibe categories → narrow to 3
2. Define visual identity: colors, fonts, style direction
3. Set content pillars: "Scene of the Day", "Process Breakdown", "Before/After"
4. Set up ComfyUI workflows for each pillar type
5. Create cron job for weekly batch generation
6. Launch with 9-post grid establishing visual consistency
7. Iterate based on engagement data