Tutorials, comparisons, and frameworks to write better prompts for ChatGPT, Claude, Gemini, Midjourney, Veo3 and Kling.
AI video fps, checked per vendor: which models fix the frame rate at 24, which let you pick it, and how LTX's fps enum shrinks your duration ceiling. Sourced and dated.
Read articleMotion strength in AI video, checked per vendor: where the dial came from, which current models still expose it, the one documented as inert, and what replaced it.
Read moreWide, medium, close, Dutch angle, negative space and 22 more composition and framing terms for AI image prompts, each defined and rendered as a copy-paste fragment.
Read moreFood photography prompts that avoid the plastic look: which angle suits which dish, hard vs diffuse light, steam and gloss cues, and copy-paste templates by shot type.
Read moreA 40-prompt AI B-roll library by editing job: establishing, cutaway, texture, transition and background-for-text, plus the aspect ratio, audio and duration settings that decide which clips cut in.
Read moreMulti shot ai video, sorted by what's real: Kling's settings.multi_shot field, Runway's multi_shot_video recipe, LTX and Veo's single-prompt techniques, and what needs separate generations.
Read moreLandscape video prompts for aerial vistas, golden-hour light and time-of-day change, with verified aspect ratio, duration and audio fields from Veo 3.1, Kling 3.0 Omni and Runway Gen-4.5.
Read moreGPT Image 2 editing and inpainting: exact mask requirements, input fidelity, multi-image edits, and what changes with no mask at all. Verified against OpenAI's own docs, Sept 2026.
Read morePerplexity Spaces don't exist under that name anymore. What Perplexity documents today as Projects, Deep Research, Model Council and Skills for recurring research, verified September 3, 2026.
Read moreWhat Black Forest Labs actually documents about FLUX.2 text rendering, which variant to use, the exact prompting method, and the limits BFL doesn't publish that Ideogram does.
Read moreWhat actually controls pacing in an AI video prompt: duration and shot counts are real, documented parameters. Slow motion and speed ramps are prompt language on every vendor checked, not a field.
Read moreClaude system prompt means three things: the API's system parameter, a claude.ai Project's custom instructions, and Claude Code's CLAUDE.md. What belongs in each, verified against Anthropic's docs.
Read moreA six-part image prompt structure (subject, setting, composition, light, style, constraints), built one piece at a time and checked against what Google, OpenAI and Ideogram document.
Read moreTen FLUX.2 photorealism prompt formulas, each built from Black Forest Labs' own documented parameters and prompting guidance, with a copy-paste template and a filled example.
Read moreNo vendor guarantees a consistent AI art style across a series. OpenAI and Google both name it as a limitation. The honest fix: a style contract, references, one sitting, and curation.
Read moreComic and manga panel prompting without naming a living artist's style: character sheets, panel-to-panel consistency, shot variety, and why lettering gets added after generation.
Read moreNo AI video vendor exposes a lens, aperture, or depth-of-field field. What Google, Kling, OpenAI, and LTX actually document instead, and the prompt language that measurably shapes focus.
Read moreStop sequences across OpenAI, Anthropic and Google, verified from each vendor's own API reference — exact field names, why OpenAI blocks stop on o3 and o4-mini, and when a schema beats it.
Read moreCamera control in AI video tools, sorted into real API fields versus prompt language only — verified from Kling, Runway, Seedance, Vidu and Veo's own documentation.
Read moreTAG is Task, Action, Goal — the shortest prompt framework in this cluster. What each field does, a worked example proving Goal matters, and the two things it's missing.
Read moreCREATE is Character, Request, Examples, Adjustments, Type of Output, Extras — six fields for a longer creative brief. Traced to its 2023 source, including where it contradicts itself.
Read moreA working system for naming, describing, and tagging saved prompts, so future-you can find the right one in seconds — not a folder system, a naming one.
Read moreA simple rule for deciding, before you start typing, whether a task deserves a saved template or is genuinely faster to just write fresh: the three-times rule, with the math behind it.
Read moreA periodic keep, merge, or retire pass for a prompt library you already have — for clutter and quality, not for prompts that broke because a model changed underneath them.
Read morePage 1 of 27