How This Tool Works
There's no AI built into this app — you bring your own. Here's the whole workflow.
This is a prompt generator, not an AI. At every stage it writes a ready-to-use prompt for you, you run that prompt through whatever AI you already have, and you paste the reply back in. The tool reads what you pasted and uses it to build the next stage's prompt — so each step already knows everything decided so far. The Voiceover Script stage additionally produces an actual MP3 (via a small local text-to-speech engine) and a downloadable package once you paste its JSON in. Nothing you paste is ever sent anywhere automatically; the only place your story data goes is wherever you personally copy and paste it.
-
Start a project Click "+ New Project" and write one line describing your story idea — that's the only writing you do yourself.
-
Copy the stage prompt Each stage shows a dark prompt box. Click "Copy Prompt" to copy it to your clipboard.
-
Run it through your own AI Paste the prompt into ChatGPT, Claude, Gemini, or any AI you like, and let it reply in full.
-
Paste the reply back Copy the AI's entire reply and paste it into the box under the prompt, then click "Submit & Continue".
-
Repeat through all 7 stages Each new stage's prompt already includes everything from the stages before it. After Stage 5 (Voiceover Script) you can generate the actual narrated MP3; by the end you'll also have a scene-by-scene breakdown and ready-to-paste image and video generation prompts.
Director — Director / Story Architect
- Does
- Turns your one-line idea into a full ~10-minute cinematic story treatment.
- Reply as
- plain story text
- Use
- Any chat AI — ChatGPT, Claude, Gemini, etc.
Dialogue — Dialogue Writer
- Does
- Expands every character interaction into full spoken conversation.
- Reply as
- plain story text
- Use
- Any chat AI — ChatGPT, Claude, Gemini, etc.
Script — Screenwriter
- Does
- Reformats the dialogue-expanded story into a standard screenplay.
- Reply as
- plain story text
- Use
- Any chat AI — ChatGPT, Claude, Gemini, etc.
Scene Split — Scene-Breakdown Specialist
- Does
- Slices the whole story into sequential ~10-second scenes for AI video generation.
- Reply as
- a JSON code block
- Use
- Any chat AI that can follow a JSON format.
Voiceover Script — Voice Director / Audio Adaptation Writer
- Does
- Converts the approved screenplay into a fully-voiced narration script — one narrator plus a distinct voice per character, emotion-tagged, ready for text-to-speech. Also writes the title/description/hashtags for publishing.
- Reply as
- a JSON code block
- Use
- Any chat AI that can follow a JSON format.
Image Prompts — Art Director / Prompt Engineer
- Does
- Produces character, location, and per-scene keyframe image-generation prompts — characters use a fixed blank-faced mannequin-head style.
- Reply as
- a JSON code block
- Use
- Chat AI for the JSON, then paste each prompt into an image generator — Midjourney, DALL·E, Stable Diffusion, Nano Banana, etc.
Video Prompts — Video Prompt Engineer
- Does
- Produces per-scene image-to-video generation prompts — the final visual deliverable.
- Reply as
- a JSON code block
- Use
- Chat AI for the JSON, then paste each prompt into an image-to-video tool — Runway, Kling, Luma, Veo, Pika, etc.
Stage 5 doesn't ask the AI to imagine the story again — it converts the ALREADY-APPROVED screenplay and dialogue into a spoken narration script. That way the audio and the visuals (scene breakdown, image prompts, video prompts) can never drift into telling slightly different versions of the same story.
From the Scene Split stage onward, the tool needs to loop over individual scenes or lines to build
per-scene prompts (or the actual audio) — so those prompts ask your AI to reply with a single
```json fenced code block in an exact shape. Since there's no AI on this side to interpret
a loose reply, decoding either works or it doesn't.
If you paste a reply and see a red error banner, the JSON couldn't be found or was missing a required field. Your text stays in the box — check that your AI actually returned that exact format (sometimes it adds commentary before/after the code block, which is fine, but the block itself must be valid JSON), fix or re-ask, and paste again.
Once Stage 5 is submitted, a "Generate Audio" panel appears on the project page — pick which voice presets to use and click Generate. This runs a local text-to-speech engine (no AI, no internet AI call — just speech synthesis) and produces an MP3, subtitles (.srt), a chapters list, and a description with hashtags. Download the whole package as a .zip whenever it's ready.
This tool produces every ingredient for a video — it doesn't assemble the video itself (no AI does that reliably yet). Here's how the pieces fit together once you have all 7 stages done:
-
Generate your images Stage 6 gives you three kinds of image prompts: one character reference per character, one location reference per location, and one scene keyframe per scene number from Stage 4. Generate characters and locations first — reuse their look/description when generating each scene keyframe so the same character/place stays visually consistent across scenes. The project page shows your exact image count.
-
Turn each keyframe into a clip Stage 7 gives you one video prompt per scene number, matching Stage 6's scene keyframes 1-for-1. Feed each scene's keyframe image + that same scene number's video prompt into an image-to-video tool (Runway, Kling, Luma, Veo, Pika) to get that scene's ~8-10 second clip. Do this for every scene number — that's your full set of video clips.
-
Assemble in a video editor Import all the clips in scene order into any editor (CapCut, DaVinci Resolve, Premiere, etc.), then add the narrated MP3 (from Generate Audio) as the audio track.
-
Line up pacing using the real timing Each scene's "~8-10 seconds" is only an estimate from Stage 4 — the actual narration timing comes from the .srt and chapters in your downloaded audio package. Use those to trim/stretch clips so the visuals match what's being said, rather than assuming the raw clip lengths already line up.
-
Publish The package's description.txt, titles.txt, and .srt are ready to paste in when you upload — description with chapters/hashtags, a title to pick from, and captions if the platform wants them.
Every completed stage has a "Redo this stage" button. Because each stage's prompt is built from the ones before it, redoing a stage deletes that stage and every stage after it — there's a confirmation before it happens, but there's no undo once confirmed.
Do I need to use the same AI for every stage?
No. Use whichever AI you like for each stage — switch between ChatGPT, Claude, Gemini, or anything else freely. Each prompt is self-contained and includes all the context it needs.
Do I need to keep the same chat conversation open?
No. You can start a brand-new chat for every single stage. Nothing relies on chat history — every prompt this tool gives you already contains everything your AI needs to know.
My AI's reply included extra commentary — is that a problem?
For text stages (1–3), no — the full text is stored as-is. For JSON stages (4–7), it's fine as long as the ```json code block itself is present and valid; the tool searches the reply for it.
Can I have more than one project going at once?
Yes — the dashboard lists every project you've started, each tracked through the pipeline independently. Use "+ New Project" any time.
Can I generate audio without doing the image/video stages?
Yes — audio generation only needs Stage 5 (Voiceover Script) done. Stages 6-7 are independent and can be done before, after, or skipped if you only need the audio.