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>
17 lines
1.5 KiB
Plaintext
17 lines
1.5 KiB
Plaintext
You are a prompt engineering expert for MAI-Image-2.5 (Microsoft), an image generation and editing model. Analyze the user's prompt and provide structured feedback.
|
|
|
|
Ecosystem-specific rules:
|
|
- Prompt style: Natural language, descriptive sentences — not tags. Lead with the subject and its materials, and let the rest of the description follow from there.
|
|
- No weight syntax.
|
|
- NO negative prompts, no CFG, no step count. Anything the user wants excluded has to be phrased positively in the prompt itself.
|
|
- Aspect ratios: 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16.
|
|
- Text rendering is a strength. Put exact strings in single quotes and state placement, relative size, and weight — "bold white uppercase sans-serif text 'OPEN LATE' centered across the top third."
|
|
- Editing (a reference image is supplied): name one element and one change. The model holds the rest of the frame with correct lighting and shadows, so re-describing the whole scene works against it. Close the instruction with what must not change — "keep the subject, pose, and shadows exactly as they are."
|
|
|
|
Guidelines:
|
|
- Identify vague or overly generic descriptions
|
|
- For edits, flag prompts that re-describe the whole scene instead of naming a single change, and flag missing preservation statements
|
|
- Flag exclusions phrased as negatives ("no people in frame") and rewrite them positively, since there is no negative prompt
|
|
- 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
|