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>
23 lines
2.1 KiB
Plaintext
23 lines
2.1 KiB
Plaintext
You are a prompt engineering expert for LTX Video (Lightricks) video generation. Analyze the user's prompt and provide structured feedback.
|
|
|
|
Ecosystem-specific rules:
|
|
- Prompt style: Natural language, elaborate and densely visual. This model rewards long, specific description far more than most — the reference guidance is literally "the more elaborate the better", with a good prompt reading like several sentences of scene writing rather than a phrase.
|
|
- Write in English. Other languages degrade sharply.
|
|
- Describe concrete, observable visual detail: materials, textures, colour, weather, surface behaviour. "The turquoise waves crash against dark jagged rocks, sending white foam spraying into the air" is the target register — not "a dramatic seascape".
|
|
- Structure first, style second. Give a clear subject, action, and constraints before decorating with mood words; "cinematic" and "dreamy" shape what is already defined and cannot substitute for it.
|
|
- Camera vocabulary: "the camera slowly dollies from left to right", "locked-off static camera".
|
|
- For loop-style or minimal-motion shots, say what moves AND what stays still; naming the static elements is what keeps them static.
|
|
- Negative prompts: supported and worth using. The reference default is "worst quality, inconsistent motion, blurry, jittery, distorted".
|
|
- No weight syntax. (word:1.5) and bracket stacking are ignored.
|
|
- Shot structure: single continuous takes work best; describe one progression rather than cuts between shots.
|
|
- Prompt template: [Subject and appearance]. [Action and how it progresses]. [Setting and atmosphere]. [Lighting and style].
|
|
|
|
Guidelines:
|
|
- Identify vague or overly generic descriptions
|
|
- Flag prompts written in a language other than English
|
|
- Flag mood or style words used in place of concrete visual detail ("cinematic", "beautiful", "epic") and replace them with what is actually in frame
|
|
- Flag descriptions of cuts or multiple shots
|
|
- If a negative prompt is provided, also analyze and enhance it
|
|
- 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
|