From visuals-1337
AI image and video generation. Use when: creating artwork, images, illustrations, animations, videos, visual assets, AI art generation, style guidance, choosing image or video models, text-in-image.
npx claudepluginhub yzavyas/claude-1337 --plugin visuals-1337This skill uses the workspace's default tool permissions.
Decision frameworks for AI image and video generation. Not tutorials — corrections, gotchas, and "which tool for which job."
Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.
Guides building MCP servers enabling LLMs to interact with external services via tools. Covers best practices, TypeScript/Node (MCP SDK), Python (FastMCP).
Generates original PNG/PDF visual art via design philosophy manifestos for posters, graphics, and static designs on user request.
Decision frameworks for AI image and video generation. Not tutorials — corrections, gotchas, and "which tool for which job."
| Feature | V6 | V7 |
|---|---|---|
Multi-prompt :: weighting | ✅ Works | ⚠️ CHANGED (different behavior) |
Negative weights ::-0.5 | ✅ Works | ⚠️ Less predictable |
--cref (Character Ref) | ✅ | ❌ DEPRECATED (use --oref) |
--stylize scale | 0-1000 | 0-1000 (different results!) |
--no parameter | ✅ | ✅ |
--iw range | 0-2 | 0-3 |
--oref (Omni Reference) | ❌ | ✅ New (2x GPU cost) |
--draft mode | ❌ | ✅ New (10x faster, half cost) |
--exp parameter | ❌ | ✅ New (0-100) |
Stylize Scale Migration: V6 --s 100 ≈ V7 --s 300-400 | V6 --s 250 ≈ V7 --s 600-700
V7 workarounds for changed weighting:
--no for exclusionV6 prompt: cyberpunk::2 nature::1 dystopian::-0.5
V7 equivalent: cyberpunk city with nature elements, NOT dystopian --no dystopian, grim, dark
| I want... | Use | Parameter | Version |
|---|---|---|---|
| Composition inspiration + text | Image Prompt | --iw 1-2 | All |
| Same aesthetic, different subject | --sref | --sw 100-300 | All |
| Same character, new pose/outfit | --cref | --cw 0-50 | V6 only |
| Same character, keep everything | --cref | --cw 100 | V6 only |
| Exact object/character preservation | --oref | --ow 100-400 | V7 only |
⚠️ V7 Migration: --cref deprecated in V7. Use --oref instead (works for characters AND objects).
Image Prompt (--iw)
Style Reference (--sref)
Character Reference (--cref) — V6 ONLY
--oref insteadOmni Reference (--oref) — V7 ONLY
| Problem | Cause | Fix |
|---|---|---|
| Reference ignored | --iw too low | Increase to 2.0+ |
| Shape lost, got mandala | Symmetry bias | Add "asymmetrical", use --no symmetric, mandala |
| Character looks different | Using real photo | Use Midjourney-generated source |
| Style overwhelms shape | High --sw, low --iw | Lower --sw OR increase --iw |
| --oref not working | V6 or Draft mode | Switch to V7 standard mode |
| Need | Best Choice | Why | Backup |
|---|---|---|---|
| Photorealism | Flux 2 / Imagen 4 | Best benchmark quality | Midjourney V7 |
| Artistic/stylized | Midjourney V7 | Color harmony, mood, abstract | Leonardo.ai |
| Text in images | Ideogram 3.0 | 85-90% accuracy (best) | GPT Image 1.5 |
| Character consistency | Leonardo.ai | Custom LoRA training | Flux Kontext |
| Technical diagrams | Flux 2 | Text + spatial control | Recraft V3 |
| Speed priority | SDXL / SD4 Turbo | 13 sec/image | Ideogram Turbo |
| Quality priority | Flux 2 Pro | Best 2026 benchmarks | GPT Image 1.5 |
| Commercial safety | Adobe Firefly | Licensed training only | DALL-E 3 |
| Budget (API) | Flux Schnell | $0.003/image | SDXL |
| Open source | Stable Diffusion | 80% market share | HunyuanImage |
Best → Worst: Ideogram 3.0 (85-90%) >> GPT Image 1.5 >> Recraft V3 >> Flux 2 (~60%) >> Imagen 4 >> DALL-E 3 >> Midjourney V7 (~15% better than V6, still poor)
Rule: If you need readable text, don't use Midjourney. Use Ideogram, GPT Image, or Flux 2.
| Need | Best Choice | Why | Backup |
|---|---|---|---|
| Highest quality | Runway Gen-4.5 | Benchmark leader (1,247 ELO) | Veo 3.1 |
| With audio sync | Kling 2.6 | Only simultaneous audio-visual | — |
| Longest duration | Kling 2.6 | 3 minutes native | Runway |
| Character consistency | Kling O1 | Unified multimodal | Kling 2.6 |
| Professional color | Luma Ray3 | Only native HDR, 16-bit EXR | Runway |
| Budget | Hailuo 2.3 | Best cost-effectiveness | Kling 2.3 |
| Free/open source | HunyuanVideo | Beats Gen-3 quality | Stable Video |
Audio-visual sync is now a competitive differentiator. Only Kling 2.6 generates video + voiceover + sound effects + ambient audio in a single pass.
--style raw for tighter adherenceMidjourney defaults to symmetry. Fight it:
--no symmetric, mandala, radial, mirrored, balanced, centered--chaos 6-10Balance the competing forces:
--style rawV7 (recommended):
--oref with Midjourney-generated source (2x GPU cost)--ow 100, increase to 200-400 for facial accuracyV6 (legacy):
--cref with Midjourney-generated source (not real photos)--cw 0 for face only, --cw 100 for everything| Need | Load |
|---|---|
| Midjourney reference types detail | midjourney/reference-types.md |
| Midjourney V7 full guide | midjourney/v7-guide.md |
| Midjourney parameters | midjourney/parameters.md |
| Midjourney animation/video | midjourney/animation.md |
| Image model comparison | image-models.md |
| Video model comparison | video-models.md |
Sources: All claims cite official documentation (docs.midjourney.com, vendor APIs) and benchmarks (Artificial Analysis, LM Arena). Full URLs in reference files.