Pick a tool. Start making.
Shot-ready prompts / Frame
Prompt Forge
Turn a half-formed idea into a full cinematic prompt: lens, lighting, camera move, grade, the lot.
Full pageStanding by
Nothing here yet
Image and video models reward specificity. Prompt Forge takes your idea and writes it back as a director would brief a DP: subject, lens, lighting, movement, grade, negative list.
Receipts
What comes out
On the right is an unedited Prompt Forge result, exactly as it lands in the panel. Dense, specific, and ready to paste into a video model without a rewrite.
Prompt Forge / output
VARIATION 1 — Flooded Neon Run A courier in a soaked black jacket leans the bike into a turn between market stalls, water fanning from the rear tyre, tarpaulin awnings dripping overhead. Shot on a 40mm anamorphic at T2, handheld follow half a metre behind the rear wheel, sodium vapour lamps and red stall signage raking across wet chrome, deep teal shadows, halation on every highlight, 35mm grain, 24fps with a slight shutter drag on the turn. NEGATIVE PROMPT clean dry ground, daylight, empty street, plastic skin, warped hands, text overlays, logo watermarks
Under the hood
Fifteen engines, one house style
Nothing here is locked to a single model. Ten engines write, five render actual video, and you swap between them mid-project. The house style holds either way: no filler, no preamble, no adjective parade.
Writing
- OpenAIGPT-4o miniHooks, ad copy, fast first passes128K context
- OpenAIGPT-4oPage copy, scripts, drop calendars128K context
- AnthropicClaude 3.5 SonnetTone work, brand voice, rewrites200K context
- AnthropicClaude 3.5 HaikuCaptions, names, quick rewrites200K context
- GoogleGemini 2.0 FlashStoryboards, long briefs, shot lists1M context
- MetaLlama 3.3 70BRaw prompts, unfiltered angles128K context
- DeepSeekDeepSeek V3Motion decks, calendars, storyboards64K context
- MistralMistral LargeAd copy, taglines, product lines128K context
- xAIGrok 2Hooks, thumbnails, edgy angles131K context
- QwenQwen 2.5 72BMultilingual copy, localisation32K context
Rendering
- ByteDanceSeedance 2.5Cinematic shots, camera moves, long takesUp to 30s / Native sound
- KuaishouKling 2.5 Turbo ProPeople, dance, product motion5s or 10s / Silent sound
- GoogleVeo 3.1 FastTalking shots, ads with sound, 4K finals4s to 8s / Native sound
- LightricksLTX-2.5 ProBroadcast cuts, inserts, b-roll6s to 10s / Native sound
- MiniMaxHailuo 02 ProQuick looks, one-line briefs, wild anglesFixed by the engine / Silent sound
Questions
Fair enough
Does this write for me or with me?
With you. Every tool takes a brief and gives back a first draft that is specific enough to argue with. The good version is the one you edit after, not the one that appears.
Why does the writing not sound like AI?
Each tool carries a house style that bans the usual tells: unleash, elevate, game-changer, seamless, delve, the lot. It also refuses to open with a summary of what it is about to do.
Which model is behind it?
Whichever one you pick. Every tool runs on the engine you choose, from GPT-4o and Claude to Gemini, Llama, DeepSeek and Grok. Swap it whenever you like and the pick follows you across every tool. Nothing you type into a brief is stored.
Can it make images or video?
It writes the prompts, it does not render the frames. Prompt Forge is built to hand a finished brief to whichever image or video model you already use.
Do I need an account?
No. Open a tool, fill the brief, take the output. There is no sign up, no saved history and no credit meter counting down in the corner.
Will more tools arrive?
Yes. The sixteen here are the ones that got used every day, split into four families: Frame, Screen, Market and Brand. Anything new has to earn its place on that grid rather than pad it out.