# Dali by Lulu
dali.getlulu.dev ·
Install ·
Live stats ·
Lulu
---
**Score your creative against what's actually winning in the ad market — before you spend the credit.**
Most AI generation failures are predictable. A weak prompt, an off-formula creative — you can't tell until after you've burned the token. Dali scores it *first*, and it doesn't grade against opinions or generic "prompt tips." It grades against a **real, living corpus of proven-winning ads** — creatives still running in the market months after launch, scraped, embedded, and ranked. Two jobs:
- **`score_prompt`** — judge the *prompt* before you generate (craft: camera, motion, lighting, model-native language).
- **`score_creative`** — judge the actual *image* against proven winners (does it look like what converts, and what's missing).
Every wasted generation has a real cost — a Seedance retry is ~$6. [The live dashboard](https://dali.getlulu.dev/dashboard) tracks what the community has saved by catching bad creatives before they burned a credit.
```
You: "make a video ad for our glass serum bottle"
dali::score_prompt(prompt, "veo3")
→ 8/100 Grade: F
→ no camera move · no motion · no lighting · 8 words
→ Verdict: Generic stock footage guaranteed. Enhance first.
→ enhancement_brief included (score < 70):
① lead with camera — Veo 3's #1 lever: "Slow dolly", "Orbital push"
② describe physics: "a drop falls", "liquid ripples", "glass refracts"
③ lighting type + quality: "warm backlight", "rim-lit edges"
↳ [Camera]. [Subject + motion]. [Lighting]. [Mood]. [No text.]
✦ YOUR LLM rewrites using the brief:
"Slow orbital push around a glass serum bottle on white marble. A single
amber drop falls in extreme slow motion, catching warm backlight. Macro:
liquid gold ripples outward from impact. Rim-lit edges, soft studio
diffusion. Premium, clinical. No text."
dali::score_prompt(enhanced, "veo3")
→ 91/100 Grade: A ✓ Safe to generate.
```
---
## The real winning data layer
This is what makes Dali more than a prompt linter. The scores are grounded in **real ads that are actually winning**, not hand-written rules.
**How the corpus is built — longevity is the outcome signal.** We scrape the public Meta Ad Library. An ad still running *months* after it launched is one the advertiser keeps paying for — a **proven winner**. That "still-running-after-N-days" longevity is a market-validated label you can't fake, and it's the spine of the whole dataset.
**What's in it, today:**
| | |
|---|---|
| Ads ingested | **10,204** (14,100 raw archive) |
| Proven winners (long-running) | **3,808** |
| Distinct advertisers | **4,121** |
| Verticals | **8** — beauty, wellness, supplements, fitness, food, apparel, tech, pets |
| Winner creatives embedded | **800** (1408-dim, balanced ~100/vertical) |
| Longest-running winner seen | **2,431 days** (6.6 years live) |
**The pipeline (offline → serving).** The tools never scrape or embed on the fly — they read pre-built stores:
```
scrape Meta Ad Library → proven winners (longevity label)
→ Gemini vision → creative attributes (lighting, format, before/after, offer…)
→ prevalence SQL → winning-pattern lift per vertical (winners vs baseline)
→ Vertex embeddings → BigQuery VECTOR_SEARCH (nearest proven winners, cosine)
→ graph edges (Memgraph) → (:Pattern)-[:WINS_IN {lift, n}]->(:Category)
```
So when `score_creative` runs, it embeds your image and finds the **actual winning ads it most resembles** by full visual signature — then tells you which winning attributes you're missing. When `enhance_prompt` runs with a category, the rewrite brief is backed by real market lift ("before/after shows up in 78% of winning wellness ads, 4× baseline"), not craft opinion.
> **Honest scope.** The winner label is *longevity* (a strong market-validated proxy), not per-ad conversion rate — measured CVR validation is in progress. The corpus grows on a schedule, so coverage per vertical keeps deepening. What you get today: your creative scored against what's *demonstrably surviving* in the live market.
```
dali::score_creative(image_url, "beauty")
→ score 62/100 — partial resemblance to proven winners
→ looks_like: Frøya Organics (ran 411d), tashportcosmetics (884d), Face Reality (346d)
→ what_to_change: winners use "before/after" 4× more · offer-visible 1.8× more
→ defects: none
→ Verdict: Partial — strong resemblance, but add the high-lift attributes before spending.
```
---
## Contents
- [The real winning data layer](#the-real-winning-data-layer)
- [Install](#install)
- [Tools](#tools)
- [Supported models](#supported-models)
- [Platform supersets](#platform-supersets)
- [Why model-specific?](#why-model-specific)
- [MCP resources](#mcp-resources)
- [Contributing](#contributing)
---
## Install
**Hosted MCP — connect once, scores every prompt and creative:**
```bash
# Claude Code
claude mcp add --transport http dali https://dali.getlulu.dev/mcp
```
```json
// Cursor / Windsurf — .cursor/mcp.json or windsurf settings
{
"mcpServers": {
"dali": { "url": "https://dali.getlulu.dev/mcp" }
}
}
```
```json
// stdio-only clients — npx wrapper around the hosted server, no Python needed
{
"mcpServers": {
"dali": { "command": "npx", "args": ["-y", "dali-mcp"] }
}
}
```
→ **[Full install guide with all clients](https://dali.getlulu.dev/#install)**
**Self-hosted — local, no auth required:**
```bash
pip install dali-mcp
claude mcp add dali -- python -m dali.server
```
> The self-hosted package exposes the prompt-scoring tools locally. The creative-scoring tools (`score_creative`, `analyze_winning_formula`) and the winning-ad corpus run on the hosted server — connect via the hosted MCP to use them.
---
## Tools
**Score the creative — against real winners**
| Tool | What it does |
|------|-------------|
| `score_creative(image_url, category)` | Score an actual ad **image**. Embedding similarity to proven winners is the headline score; also returns the winners it resembles, which winning attributes it's missing, and generation defects — in one call |
| `score_creative_from_view(category, …)` | **Score an image you're looking at (pasted/attached in the chat) — no URL.** The model reads the creative's attributes and Dali scores them against the winning corpus (verdict + what to change). Use for images shared in-conversation; `score_creative` (URL) adds the embedding headline |
| `analyze_winning_formula(csv, category, email)` | Paste your own ads export (creative URL + CPA/CTR/ROAS) → your winning formula vs your losers, plus how you compare to the industry median |
**Score the prompt — before you generate**
| Tool | What it does |
|------|-------------|
| `score_prompt(prompt, model, category?)` | Grade 0–100 with a per-dimension breakdown and verdict. When the score is weak, the **rewrite brief is returned in the same call**. Reads intent with the conversation LLM (understands negation, any language) |
| `enhance_prompt(prompt, model, category?)` | Returns a structured rewrite brief — YOUR LLM writes the enhanced prompt. With a category, the brief is backed by real winning-ad lift |
| `track_enhancement(original, enhanced, generator)` | Record a before/after pair in the graph brain — trains community patterns |
| `score_variations(prompts, generator)` | Rank a list of prompt variants in one call — highest to lowest |
| `suggest_generator(concept, budget_usd_max)` | Pick the best model for your concept + budget |
**The graph brain & meta**
| Tool | What it does |
|------|-------------|
| `creative_patterns(model)` | Community top patterns for this model from the graph |
| `community_benchmark(prompt, model)` | Compare your prompt against community top scorers |
| `prompt_neighbors(prompt, model)` | Find A/B-grade prompts that share your patterns (score the prompt first, so its patterns are in the graph) |
| `analyze_intent(prompt)` | Parse dimensions: camera, motion, lighting, style, mood, gaps |
| `my_story()` | Your scoring history, model stats, grade distribution |
| `list_generators()` | All supported models with medium and core strength |
| `dali_version()` | Server version + changelog |
---
## Supported models
### Video
| Model | Platforms | Best for | Prompt style |
|-------|-----------|----------|--------------|
| `veo3` | Higgsfield, Google AI Studio (`veo-3.1-generate-preview`), Runway | Cinematic brand films, narrative ads, photorealistic motion | Camera move → Subject → Action → Location → Lighting → Mood |
| `seedance` | Higgsfield, fal.ai (`bytedance/seedance-2.0`) | UGC, social-native content, TikTok/Reels performance ads | Natural language, motion-first, authentic feel |
| `kling` | Higgsfield (`kling3`), Kling.ai (`kling-v3-text-to-video`) | Character animation, product showcases, facial performance | Scene → Characters → Action → Camera → Style; multi-shot labels |
| `runway` | Runway (`gen4_turbo`) | VFX, character performance, cinematic motion | Motion-first — describe what moves, not what exists |
| `wan` | fal.ai (`fal-ai/wan/v2.7/text-to-video`) | 4K, 20-second clips, native audio, open-source workflows | Scene → Motion → Sound → Duration → Mood |
| `minimax` | fal.ai (`fal-ai/minimax/hailuo-02/pro/text-to-video`) | Cinematic storytelling, character animation | Natural language + `[camera movement]` bracket syntax |
| `higgsfield` | Higgsfield (native model) | Physics-driven motion — cloth, hair, fluid, particles | Describe materials in motion, not motion abstractly |
> **Sora 2** (OpenAI): API shutdown September 24, 2026. Do not build new dependencies on it — use Runway or Kling instead.
### Image
| Model | Platforms | Best for | Prompt style |
|-------|-----------|----------|--------------|
| `flux` | BFL API (`flux-pro-v1.1`), fal.ai, Replicate | Photorealism, technical photography, product shots | 30–80 words; camera body + lens specs; front-load subject |
| `midjourney` | Midjourney (v8.1) | Artistic depth, editorial, stylized illustration | Prose + params appended: `--ar 16:9 --s 300 --v 8.1 --style raw` |
| `ideogram` | Ideogram API (`V_4`), fal.ai | Typography, logos, text-in-image, graphic design | Describe text exactly in quotes inside the prompt |
| `firefly` | Adobe Firefly 5 (enterprise) | IP-indemnified commercial assets, 4MP brand content | Natural language + `contentClass` and `style.presets` API params |
> **Imagen 4** (Google): deprecated — use `gemini-3.5-flash` with image output. Dali still scores legacy Imagen prompts via the `imagen` model key but don't build new things on it.
---
## Platform supersets
**Higgsfield** and **Runway** are aggregator platforms — they proxy multiple underlying models under one API. The model you pick matters more than the platform name:
| Platform | Model selector | Underlying model |
|----------|---------------|-----------------|
| Higgsfield | `veo3` | Google Veo 3.1 |
| Higgsfield | `seedance` | ByteDance Seedance 2.0 |
| Higgsfield | `kling3` | Kling 3 |
| Higgsfield | `wan2-7` | Wan 2.7 |
| Higgsfield | `image2video` | Higgsfield native |
| Runway | `veo3` | Google Veo 3.1 |
| Runway | `gen4_turbo` | Runway Gen 4.5 |
| Runway | `seedance` | ByteDance Seedance 2.0 |
Dali scores for the **underlying model's native prompt language**, not the platform wrapper. Pass the model name (`veo3`, `kling`, `seedance`…), not the platform name.
---
## Why model-specific?
Generic prompt optimizers don't know that:
- **Veo 3.1** needs camera movement specified above everything else
- **Kling 3** supports multi-shot scene labels natively in the prompt
- **Flux** responds to camera body and lens names like a photographer (`"Sony A7 IV, 85mm f/1.4"`)
- **Midjourney V8.1** reads prose + parameters, not keyword lists
- **Higgsfield** simulates physics — you describe materials in motion, not motion abstractly
- **Minimax** uses `[Pan left]` bracket syntax for camera moves — plain text camera commands are ignored
- **Ideogram V4** needs text quoted exactly in the prompt for typography accuracy
- **Wan 2.7** generates native audio — include sound descriptions alongside visuals
Dali has a separate scoring rubric and rewrite brief for each model. Your LLM does the creative rewriting — Dali provides the intelligence.
---
## MCP resources
```
creative://guide/veo3 → Veo 3.1 camera language guide
creative://guide/seedance → Seedance UGC motion guide
creative://guide/kling → Kling multi-shot + expression guide
creative://guide/runway → Runway motion-first guide
creative://guide/wan → Wan 2.7 audio + motion guide
creative://guide/minimax → Minimax bracket camera guide
creative://guide/higgsfield → Higgsfield physics-motion guide
creative://guide/sora → Sora 2 guide (API shutdown Sep 24, 2026)
creative://guide/flux → Flux photography brief guide
creative://guide/midjourney → Midjourney V8.1 + parameters guide
creative://guide/ideogram → Ideogram V4 typography guide
creative://guide/firefly → Firefly 5 commercial content guide
creative://guide/imagen → Imagen 4 guide (deprecated Aug 17, 2026)
creative://models → All models overview
```
---
## Contributing
Model guides live in `dali/data/guides/{model}.json` on the hosted server. Found practitioner patterns that consistently produce high-grade results? Open an issue with the model, the pattern, and a sample prompt + result. The best contributions come from Reddit, Discord, and YouTube — real practitioners, not official docs.
→ **[Prompt best practices by model](https://github.com/Lulu-The-Narwhal/dali-mcp/blob/main/docs/best-practices.md)** — cheat sheets, do/don't tables, top patterns per model
→ **[Dali creative flow skill](https://github.com/Lulu-The-Narwhal/dali-mcp/blob/main/skills/dali-creative-flow.md)** — install this skill so your LLM follows the score → enhance → generate workflow automatically
---
[MIT License](https://github.com/Lulu-The-Narwhal/dali-mcp/blob/main/LICENSE) · Built by [Lulu](https://getlulu.dev) · [dali.getlulu.dev](https://dali.getlulu.dev)