$npx -y skills add prime-skills/runcomfy-agent-skills --skill flux-2-kleinGenerate images with Flux 2 Klein (Black Forest Labs' distilled fast variant of Flux 2) on RunComfy — bundled with the model's documented prompting patterns so the skill gets sharper output than naive prompting against the same model. Documents Flux 2 Klein's strengths (sub-secon
| 1 | # Flux 2 Klein — Pro Pack on RunComfy |
| 2 | |
| 3 | [runcomfy.com](https://www.runcomfy.com/?utm_source=skills.sh&utm_medium=skill&utm_campaign=flux-2-klein) · [9B model](https://www.runcomfy.com/models/blackforestlabs/flux-2-klein/9b/text-to-image?utm_source=skills.sh&utm_medium=skill&utm_campaign=flux-2-klein) · [4B model](https://www.runcomfy.com/models/blackforestlabs/flux-2-klein/4b/text-to-image?utm_source=skills.sh&utm_medium=skill&utm_campaign=flux-2-klein) · [GitHub](https://github.com/agentspace-so/runcomfy-skills/tree/main/flux-2-klein) |
| 4 | |
| 5 | Black Forest Labs' **Flux 2 Klein** (the distilled, low-latency variant of Flux 2) hosted on the **RunComfy Model API** — no API key, async REST. |
| 6 | |
| 7 | ```bash |
| 8 | npx skills add agentspace-so/runcomfy-skills --skill flux-2-klein -g |
| 9 | ``` |
| 10 | |
| 11 | ## When to pick this model (vs siblings) |
| 12 | |
| 13 | Flux 2 Klein's distinct strength is **latency-first creative iteration**: sub-second feedback enables live art-direction sessions and rapid product visualization that batch-style models can't sustain. Pick it when **iteration speed matters more than ceiling resolution**. |
| 14 | |
| 15 | | You want | Use | |
| 16 | |---|---| |
| 17 | | Real-time / live art-direction sessions | **Flux 2 Klein 4B** | |
| 18 | | Fast iteration with strong detail at the end | **Flux 2 Klein 9B** | |
| 19 | | Multi-reference brand styling with consistent looks | **Flux 2 Klein** | |
| 20 | | 2K–4K hero images, max resolution | Seedream 5 | |
| 21 | | Maximum prompt adherence + extreme detail | Flux 2 Pro | |
| 22 | | Embedded text, logos, multilingual signage | GPT Image 2 | |
| 23 | | Hyperrealistic portrait | Nano Banana Pro | |
| 24 | |
| 25 | If the user said "Flux 2 Klein" / "BFL Klein" / "flux klein" explicitly, route here regardless. If they said "Flux 2" generically, ask whether they want **Klein** (fast) or **Pro** (max quality) before defaulting. |
| 26 | |
| 27 | ## Prerequisites |
| 28 | |
| 29 | 1. **RunComfy CLI** — `npm i -g @runcomfy/cli` |
| 30 | 2. **RunComfy account** — `runcomfy login` opens a browser device-code flow. |
| 31 | 3. **CI / containers** — set `RUNCOMFY_TOKEN=<token>` instead of `runcomfy login`. |
| 32 | |
| 33 | ## Endpoints + input schema |
| 34 | |
| 35 | Two variants, same endpoint shape, same prompt grammar. |
| 36 | |
| 37 | ### `blackforestlabs/flux-2-klein/9b/text-to-image` |
| 38 | |
| 39 | The fidelity-first variant. Use for polish / final output. |
| 40 | |
| 41 | | Field | Type | Required | Default | Notes | |
| 42 | |---|---|---|---|---| |
| 43 | | `prompt` | string | yes | — | Up to ~512 tokens. Longer degrades. | |
| 44 | | `steps` | int | no | 25 | 4–50. **Step-distilled architecture** — 4–8 enough for concepting; ~25 for polish; >25 buys little. | |
| 45 | | `width` | int | no | 1024 | 512–1536 typical. **Aspect ratio capped at 16:9**, max ~2K total. | |
| 46 | | `height` | int | no | 1024 | Match `width`'s aspect intent. | |
| 47 | |
| 48 | ### `blackforestlabs/flux-2-klein/4b/text-to-image` |
| 49 | |
| 50 | The latency-first variant. Sub-second 4-step inference. Use for live iteration / concepting. |
| 51 | |
| 52 | Same field set as 9B. Default `steps` is effectively 4 — the variant is built for that step count. |
| 53 | |
| 54 | ### Reference images (both variants) |
| 55 | |
| 56 | Up to **4 simultaneous reference images** are supported on the same endpoint for style transfer / guided composition. The exact field name in the JSON body is documented on the [model's API tab](https://www.runcomfy.com/models/blackforestlabs/flux-2-klein/9b/text-to-image?utm_source=skills.sh&utm_medium=skill&utm_campaign=flux-2-klein) — pass it through the CLI verbatim. Reference-image use enables editing-style workflows without a separate `/edit` endpoint. |
| 57 | |
| 58 | ## How to invoke |
| 59 | |
| 60 | **Fast concepting (4B, sub-second):** |
| 61 | |
| 62 | ```bash |
| 63 | runcomfy run blackforestlabs/flux-2-klein/4b/text-to-image \ |
| 64 | --input '{"prompt": "<user prompt>"}' \ |
| 65 | --output-dir <absolute/path> |
| 66 | ``` |
| 67 | |
| 68 | **Polish / final (9B, ~25 steps):** |
| 69 | |
| 70 | ```bash |
| 71 | runcomfy run blackforestlabs/flux-2-klein/9b/text-to-image \ |
| 72 | --input '{ |
| 73 | "prompt": "<user prompt>", |
| 74 | "steps": 25, |
| 75 | "width": 1024, |
| 76 | "height": 1024 |
| 77 | }' \ |
| 78 | --output-dir <absolute/path> |
| 79 | ``` |
| 80 | |
| 81 | **Wide-format poster:** |
| 82 | |
| 83 | ```bash |
| 84 | runcomfy run blackforestlabs/flux-2-klein/9b/text-to-image \ |
| 85 | --input '{"prompt": "<user prompt>", "width": 1536, "height": 864}' \ |
| 86 | --output-dir <absolute/path> |
| 87 | ``` |
| 88 | |
| 89 | The CLI submits, polls every 2s until terminal, then downlo |