Run any awesome-gpt-image-2 or Seedance prompt as a one-line Claude Code command.
/headshot · /dance · /poster · 848 cases · pay-per-image USDC on Base via BlockRun
Other awesome lists tell you what prompt to copy-paste. This one runs them for you — directly from Claude Code, with output that drops into your project folder, billed by the call (no subscription).
# One-line install (registers MCP + clones bundle + symlinks /headshot, /dance, /poster)
curl -fsSL https://raw.githubusercontent.com/BlockRunAI/Claude-Code-GPT-IMAGE2-SeeDance-BlockRun/main/install.sh | bashPrefer not to pipe to bash? Manual two-step lives in INSTALL.md.
Restart Claude Code, fund your BlockRun wallet on first run
(> top up my blockrun wallet — takes ~60 seconds), and:
> /headshot ./me.jpg
> /dance ./me.jpg --style hiphop
> /poster "Last Light" --genre thriller
That's it.
The thumbnails below are real outputs the underlying models produce — served directly from the source
awesome-*repos (raw URLs, no copies stored here). Click through to open the full case file with the prompt, attribution, and the X/Twitter post it came from.
| /headshot ~$0.12 · ~10s |
/dance ~$0.75 · 60–180s |
/poster ~$0.12 · ~15s |
 |
 |
 |
| Selfie → studio portrait. 5 styles built-in (corporate, creative, startup, actor, linkedin-2025). | Photo → 5-second video. 7 choreography presets including terracotta-disco 兵马俑迪斯科. |
Title → cinema-grade key art. 8 genres, multilingual title typography. |
Five gold-mine awesome lists (EvoLinkAI, freestylefly, EvoLinkAI/awesome-seedance-2-guide, ZeroLu/awesome-seedance, YouMind-OpenLab/awesome-seedance-2-prompts) collectively curate ~850 viral GPT Image 2 + Seedance prompts. But they're static — you copy a prompt, paste it into someone's chat UI, fight with the upload, and pay through some subscription you never wanted.
This bundle turns them into executable Claude Code commands:
- What you install: one line of
curl … install.sh | bash— it registers the BlockRun MCP server (claude mcp add blockrun … npx @blockrun/mcp@latest), clones the bundle to~/.claude/blockrun-art-bundle/, and symlinks the three skills into~/.claude/skills/so they show up as bare/headshot,/dance,/poster. - What it does: thin, polished slash commands that pick the right model, build a vetted prompt template, call the right MCP tool, and drop a file in your project directory. Three commands at v1, more coming.
- Underneath: the BlockRun gateway routes to OpenAI, ByteDance, xAI, Google, and Z.ai — pay per call in Base USDC via x402. No API keys, no subscriptions, no minimums.
You type /headshot ./me.jpg --style startup
↓
[this plugin] picks template + variables, calls
↓
mcp__blockrun__blockrun_image ← from `@blockrun/mcp` (registered via claude mcp add)
↓
BlockRun gateway · x402 USDC settle · OpenAI gpt-image-2
↓
You get ./blockrun-out/2026-05-01T143200Z-headshot/headshot.png
Already have BlockRun MCP registered? Skip Step 1 —
claude mcp addis idempotent at the same name and a re-add is harmless.
/headshot ./me.jpg
/headshot ./me.jpg --style startup
/headshot ./me.jpg --style actor
/headshot ./me.jpg --all # 4 styles in one go
Five built-in styles: corporate (default), creative, startup,
actor, linkedin-2025. Source identity is preserved 1:1 — only
wardrobe, lighting, and background are restyled.
| Param | Default | Notes |
|---|---|---|
image |
required | JPG / PNG / HEIC / WebP, local path or URL |
--style |
corporate |
or creative, startup, actor, linkedin-2025, all |
| Cost | ~$0.12 | per style; --all = ~$0.48 |
Full prompt templates: prompts/headshot.md ·
Skill: skills/headshot/SKILL.md
/dance ./me.jpg
/dance ./me.jpg --style hiphop
/dance ./me.jpg --style terracotta-disco # 兵马俑迪斯科
/dance ./me.jpg --style freestyle-from-music \
--music "lo-fi jazz 75 bpm"
/dance ./me.jpg --duration_seconds 8 --style kpop
Six choreography presets: hiphop, ballet, contemporary, kpop,
terracotta-disco, tiktok-trend, plus freestyle-from-music for
custom soundtracks. Powered by ByteDance Seedance 2.0 Fast (the
price/quality knee — see docs/COSTS.md).
Outputs both dance.mp4 and a shareable looping dance.gif (if ffmpeg
is installed locally).
| Param | Default | Notes |
|---|---|---|
image |
required | full-body or 3/4-body works best |
--style |
tiktok-trend |
one of 7 presets |
--duration_seconds |
5 |
range 1–10 |
--music |
— | only for freestyle-from-music |
| Cost | ~$0.75 | for default 5s; $0.15/sec |
Full prompt templates: prompts/dance.md ·
Skill: skills/dance/SKILL.md
/poster "Last Light" --genre thriller
/poster "Founders" --genre documentary --tagline "they bet everything"
/poster "BlockRun Live" --genre event --accent_color "#0066ff" --portrait
/poster "兵马俑：复活" --genre sci-fi
/poster "Coffee with the CEO" --genre documentary --square # podcast cover
Eight genre presets: thriller, sci-fi, romcom, documentary,
concert, horror, kids-animation, event. gpt-image-2's killer
feature — multilingual title typography that actually reads correctly,
including CJK / Arabic / Cyrillic.
| Param | Default | Notes |
|---|---|---|
title |
required | rendered exactly as typed, multilingual safe |
--genre |
documentary |
one of 8 presets |
--tagline, --subject, --credits |
— | optional polish |
--landscape / --portrait / --square |
landscape | aspect (1792×1024 / 1024×1792 / 1024×1024) |
| Cost | ~$0.12 | per call |
Full prompt templates: prompts/poster.md ·
Skill: skills/poster/SKILL.md
| Command | Cost | Wall time |
|---|---|---|
/headshot (1 style) |
~$0.12 | ~10 s |
/headshot --all (4 styles) |
~$0.48 | ~40 s |
/dance (5 s, default) |
~$0.75 | 60–180 s |
/dance (10 s, max) |
~$1.50 | 100–240 s |
/poster (any aspect) |
~$0.12 | ~15 s |
No subscription. No charge if a call times out or fails. Settled on
Base mainnet USDC via x402. The wallet's private key never leaves
~/.blockrun/.session — only signed payment authorizations travel.
A typical creator session ($5 USDC top-up) covers ~40 headshots, or 6 dance
videos, or 40 posters, or any mix. See docs/COSTS.md
for the full breakdown of all underlying gateway prices.
All 60 thumbnails below are real, click-through demos rendered directly from the source
awesome-*repos via raw URLs (no copies stored here, repo stays small). Click any thumbnail to open the case file (full prompt, model, attribution, X/Twitter source post).
Want the full 800+ case catalog? See prompts/case-library/INDEX.md — grouped by tag and source repo.
| Workflow | Cases |
|---|---|
| Text → image | 259 |
| Text → video | 227 |
| Image → image (edit) | 328 |
| Image → video | 34 |
| Total | 848 unique cases (94% with hero image attached) |
By source repo:
| Source | Cases | Workflows |
|---|---|---|
EvoLinkAI/awesome-gpt-image-2-prompts |
300 | text→image, image→image |
freestylefly/awesome-gpt-image-2 |
352 | image→image edits |
EvoLinkAI/awesome-seedance-2-guide |
63 | image→video |
ZeroLu/awesome-seedance |
28 | text→video (cinematic) |
YouMind-OpenLab/awesome-seedance-2-prompts |
105 | text→video |
Every case is a single markdown file with normalized frontmatter (title,
source, credit, workflow, model, tags) plus the original demo image
embedded. See prompts/case-library/INDEX.md
for the full browseable catalog grouped by tag and source.
v1 ships with 3 polished slash commands (
/headshot,/dance,/poster). The 800+ cases above are the runway for v1.1+ — each tag in the index is a candidate slash command in waiting.
| Vibe | Source | One-line command |
|---|---|---|
| Studio Ghibli portrait | awesome-gpt-image-2 | /headshot ./me.jpg --style ghibli (v1.1) |
| 9-grid character sheet | awesome-gpt-image-2-prompts | /character-sheet "Hua, terracotta scout" (v1.1) |
| Glassmorphism UI mock | awesome-gpt-image-2 | /ui-system social --style glass (v1.1) |
| Movie poster (thriller) | this bundle | /poster "Last Light" --genre thriller |
| 5-second dance loop | awesome-seedance-2-guide | /dance ./me.jpg --style tiktok-trend |
| K-pop stage clip | awesome-seedance-2-guide | /dance ./me.jpg --style kpop |
| LinkedIn "promoted" headshot | awesome-gpt-image-2-prompts | /headshot ./me.jpg --style linkedin-2025 |
| Podcast cover (square) | this bundle | /poster "Coffee with the CEO" --square |
| ... | ... | ... |
- Native CC integration. No browser, no copy-paste, no separate tools. Slash commands inside the editor you already use.
- Pay per call. $0.12 here, $0.75 there. Stop when you're done. Top up when you want. No "10 free generations a month, then $20/mo".
- Multi-model under one roof. GPT Image 2 for typography. Seedance for video. DALL-E 3 / Nano Banana / Grok / CogView available too — switch with one parameter.
- Crypto-native, but invisible. Base USDC is the rails, but you only see USD prices. Your wallet auto-creates on first use. Top up with any wallet that speaks Base.
- Deterministic outputs. Files land in your project directory. Easy to commit (don't), share, post, or feed into the next stage of your pipeline.
| Version | Commands |
|---|---|
| v1.0 (this release) | /headshot, /dance, /poster |
| v1.1 | /character-sheet, /ui-system, /ghibli |
| v1.2 | /lookbook, /ad-series, /card-deck, /unbox |
| v2.0 | /blockrun-art generate "<free description>" smart router; full case-library executable as discoverable subcommands |
Want to influence what comes next? Open an issue or vote in the discussions: github.com/BlockRunAI/Claude-Code-GPT-IMAGE2-SeeDance-BlockRun.
The viral prompt patterns in prompts/case-library/ were curated from:
- EvoLinkAI/awesome-gpt-image-2-prompts — 300 cases · text-to-image and image-to-image template gold mine
- freestylefly/awesome-gpt-image-2 — 352 cases · image-to-image edits, style transfers, character consistency
- EvoLinkAI/awesome-seedance-2-guide — 63 cases · Seedance text-to-video and image-to-video case studies
- ZeroLu/awesome-seedance — 28 cases · cinematic Seedance 2.0 prompts (Hollywood, Denis Villeneuve, Wong Kar-wai-style multi-shot scripts)
- YouMind-OpenLab/awesome-seedance-2-prompts — 105 cases · structured Seedance 2.0 prompts with detailed scene/shot specifications and Cloudflare Stream video previews
Each case file under prompts/case-library/from-*/ includes
source_url and credit in its frontmatter — go give those repos a star.
The underlying gateway is BlockRun.ai — x402 micropayments on Base, no subscriptions, no API keys.
MIT — see LICENSE. Cases under prompts/case-library/from-*/
retain the licenses of their original repositories (all five sources are
CC BY 4.0 or MIT-compatible at the time of this release).
SEO keywords (for indexing)
claude code, claude code skill, claude code plugin, claude code marketplace, gpt-image-2, gpt image 2, openai gpt image, dall-e 3, nano banana, seedance, seedance 2.0, seedance 1.5, bytedance video, bytedance ai video, blockrun, blockrun.ai, blockrun mcp, x402, x402 micropayments, USDC base, base mainnet, base usdc payments, ai image generation, ai image generator, ai video generation, ai video generator, image to video, img2vid, image-to-image edit, photo to video, headshot generator, ai headshot, professional headshot ai, linkedin photo ai, movie poster generator, ai movie poster, film poster ai, key art ai, dance video generator, ai dance video, tiktok dance ai, viral dance video, viral ai prompts, awesome gpt image 2, awesome seedance, mcp server, model context protocol, claude mcp, character sheet ai, ui mockup ai, ghibli style ai, kpop video ai