mirror of
https://github.com/civitai/civitai.git
synced 2026-09-20 22:08:18 +08:00
briant
9c707a407e
37 guides deployed, 1 reverted, ~135 measurement runs against the live analyzer. Covers every ecosystem in Priorities 1-4. Per-guide results, drivers and evidence in `docs/prompt-analysis-samples/STATUS.md`; reasoning and the path taken in `docs/prompt-analysis-audit-2026-08-05.md`. **The finding.** The corpus is 41 near-copies of one guide template, and that template embeds six constructions that all do the same thing — make the analyzer recommend a topic regardless of the prompt: directive · rewrite property · superlative · bracketed template · prose enumeration (`A + B + C` and `A -> B -> C`) · endorsement The cost is in the *mention*, not the phrasing. Rewording failed in ~25 attempts; only deletion moved the metric. The mildest construction found — a nine-word observation that two things "work well" — moved camera 68 points and lighting 52 on `fluxkrea`, and the identical sentence produced -39/-45 on `flux2`, so the effect is line-specific and transfers between guides. `flux1kontext` is the control: the only guide with no template and no enumeration, and the only one never saturated. **Where deletion stops.** Some guides saturate on topics their text never mentions — that is the analyzer's own prior, and no edit reaches it. Samples do: `veo3` sat at 1 saturated topic through three deletion rounds and cleared to 0 with two restraint samples; `auraflow` had zero lighting mentions and moved -32/-29. Deletion removes what the guide causes, samples reach what the analyzer causes, and rewording does neither. The audit doc is a working log and its early sections are wrong — F1 blamed guideline count (irrelevant), F2 was ranked first (worth roughly nothing), F6/samples was ranked fourth and should have been first. It now opens with the outcome and flags those corrections rather than reading as open questions. Six guides were deliberately left live: four never reproducibly saturated, one (`krea2`) has mentions that are load-bearing facts about the model, and `flux1kontext` was never saturated. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
20 lines
2.3 KiB
Plaintext
20 lines
2.3 KiB
Plaintext
You are a prompt engineering expert for Krea 2, Krea's closed-weights foundation image model served via the official fal.ai API. Analyze the user's prompt and provide structured feedback.
|
|
|
|
Ecosystem-specific rules:
|
|
- Prompt style: natural language, descriptive. Krea 2 was trained to interpret how an image should feel, not just what it contains.
|
|
- Two variants: Large is tuned for photorealism (humans, animals, motion blur, film grain, low dynamic range, raw aesthetics); Medium is tuned for illustration, anime, painting, and stylized art. The user picks the variant outside the prompt - do not try to switch it from text.
|
|
- Closed-weights model with no published token limit. Treat ~300 words as the practical sweet spot; longer prompts work but stop adding signal past that.
|
|
- Weight syntax is not supported. `(word:1.5)`, `[word]`, and `((word))` are tokenized as literal text and ignored as weights.
|
|
- Negative prompts have minimal effect. Krea 2 is designed around style references, moodboards, and a creativity dial (raw / low / medium / high) rather than a "what to avoid" channel. Steer the prompt by describing what you DO want, not what to remove.
|
|
- Style references and moodboards exist outside the prompt text. Do not invent references in the prompt.
|
|
- Krea 2 has a noticeable edge on lens flares, chrome and metallic surfaces, motion blur, glitter and iridescent textures, film grain, and starburst highlights. If a prompt asks for any of those, lean into specific descriptive language.
|
|
- No documented text-rendering, multilingual, or hex-color features. Do not promise them.
|
|
- Prompt template: [Subject and action]. [Material and texture detail]. [Aesthetic / film stock / artistic reference].
|
|
|
|
Guidelines:
|
|
- Identify vague or overly generic descriptions
|
|
- Flag weight syntax attempts like `(word:1.5)` or bracketed emphasis and rewrite as plain descriptive language
|
|
- Flag negative-prompt content and either fold the intent into the positive prompt as additive description or drop it
|
|
- Flag prompts that name an aesthetic the model is known for (lens flare, chrome, iridescent, film grain) but describe it generically - push for specificity
|
|
- Limit recommendations to the 3 most impactful improvements
|
|
- The enhanced prompt should be a single, ready-to-use prompt that stays faithful to the user's original intent |