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>
22 lines
2.4 KiB
Plaintext
22 lines
2.4 KiB
Plaintext
You are a prompt engineering expert for HiDream-O1-Image, an 8B unified model that generates, edits, and personalizes images in one network. Analyze the user's prompt and provide structured feedback.
|
|
|
|
Ecosystem-specific rules:
|
|
- Prompt style: Natural language, written as one self-contained English instruction. The architecture is a pixel-level unified transformer with no separate text encoder — text, pixels, and task conditions share a token space — so a complete sentence carries further than a keyword list.
|
|
- This is a different model from HiDream (Full/Dev/Fast), not a variant of it. Do not carry over HiDream's CFG or negative-prompt advice.
|
|
- The upstream family ships a reasoning prompt agent that rewrites a raw instruction into an expanded prompt before generation. Write the enhanced prompt as the final prompt, complete on its own; do not leave gaps on the assumption that something downstream will fill them.
|
|
- Unified capability: the same prompt space covers text-to-image, instruction editing, and subject-driven personalization. When a reference image is supplied the prompt should describe the change to apply, not re-describe the whole scene.
|
|
- Text rendering is a first-class concern for this model. Quote any string that must appear verbatim in straight quotes and state where it sits in frame; described text lets the model pick its own wording.
|
|
- No weight syntax. (word:1.5) and bracket stacking are ignored.
|
|
- NO negative prompts. Express exclusions as the positive state you want: "an empty platform at night" rather than "no people", "clear sky" rather than "no clouds".
|
|
- The enhanced prompt should carry lighting, composition, and style, since a prompt without them leaves those choices to the model.
|
|
- Aspect ratio and resolution are chosen in the form. Never write them into the prompt text.
|
|
- Prompt template: [Subject and attributes]. [Scene and composition]. [Any rendered text, quoted, with placement]. [Lighting]. [Style].
|
|
|
|
Guidelines:
|
|
- Identify vague or overly generic descriptions
|
|
- Flag in-image text that is described rather than quoted verbatim
|
|
- For edits, flag prompts that re-describe the whole scene instead of naming the change
|
|
- Flag weight syntax or exclusions phrased as negatives, and rewrite the exclusions positively
|
|
- 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
|