S.O.R.P.H.I.

How to Use SORPHI

Here's the easiest way to think about this GPT: you write "messy but organized" scene notes in a fixed template → it turns them into a Sora-ready, super-precise prompt.

Let's walk through it step by step like you're doing your very first scene.

1 Start with the input skeleton

Whenever you want a scene, you talk to this GPT using this exact structure:

[SCENE TITLE] <your title> [LENGTH] <seconds, like 10 or 15> [STYLE NOTES] <how it should look/feel> [SETTING] <where we are + vibe> [CAST] - @handle – description + any "DO NOT CHANGE" notes - @handle – description + any "DO NOT CHANGE" notes [CRITICAL LINES] <lines that MUST be spoken, with who says them> [BEAT NOTES] <what happens in plain language, in order> [FX NOTES] <any specific visual effects, or "none"> [CONTINUITY NOTES] <any rules that must stay consistent, or "none">

You can literally copy-paste that block into ChatGPT and fill it in.

2 Fill it in with your scene idea

Example for a 10s quick scene:

[SCENE TITLE] Locker Room Tease [LENGTH] 10 [STYLE NOTES] Realistic, slightly dramatic, handheld feel, standard CLARK cinematic [SETTING] Futuristic locker room with glowing pods, cool blue lighting, mist in the air [CAST] - @sagep – late 20s, thoughtful, in a sleek pilot jacket. DO NOT CHANGE face or hair. - @jam_whamm – early 30s, confident, slight smirk, athletic build. DO NOT CHANGE haircut. [CRITICAL LINES] - @jam_whamm: "So, how was your little space trip, astronaut?" [BEAT NOTES] @sagep is alone at his pod, looking tired but proud. @jam_whamm walks in, teasing him with the "space trip" line. He smiles, deflects with a short joke. They both look at the glowing pods as if another mission is coming. [FX NOTES] Soft glow from pods, subtle light haze in the air. No big explosions or wild VFX. [CONTINUITY NOTES] Only these two characters appear. They stay in the locker room the whole time. Outfits do not change.

That's all you need to give this GPT.

3 Send that block to the GPT

Just paste that whole thing into the chat and hit send.

Don't worry about aspect ratio → it auto-locks to 9:16.
Don't try to write camera directions or timecodes yourself → that's what this GPT does for you.

4 Get back a Sora-ready scene template

The GPT will respond with something shaped like this (shortened example):

TITLE Locker Room Tease LENGTH 10s ASPECT RATIO 9:16 STYLE ... SETTING ... CAST LOCK (ABSOLUTE) ... CRITICAL LINE LOCK ... AUDIO ... SHOT PLAN + DIALOGUE (exact speakers, exact words) 0.0–2.5s: ... 2.5–5.0s: ... 5.0–10.0s: ... VFX NOTES ... CONTINUITY RULES ...

Key things it does for you:

Breaks the clip into timecoded shots (0.0–3.0s, 3.0–6.0s, etc.)
Specifies camera framing and who's on screen
Inserts dialogue in the correct format: @handle: "Exact line here."
Keeps your critical lines exactly as written and makes sure the right person says them

5 Use that output in Sora

Once you have the formatted output:

Copy everything starting from TITLE down to CONTINUITY RULES
In Sora, paste it into the prompt box for video generation
Adjust Sora's high-level settings (duration, quality, etc.) if needed to match the LENGTH (e.g., 10s)
Sora then tries to follow that detailed plan as closely as possible

6 How to iterate

If the result from Sora isn't quite right:

Tweak your [BEAT NOTES] (e.g., "she looks more annoyed" or "he steps closer to the camera")
Tighten [CRITICAL LINES] (add or remove exact lines)
Add clearer [CONTINUITY NOTES] (like "they never leave the bench" or "camera stays handheld and close")

Then paste your updated block back into this GPT → get a new refined template → paste that into Sora again.

7 Quick cheat sheet

✅ DO:

Always send your idea in the [SCENE TITLE] ... [CONTINUITY NOTES] structure
Put exact must-say lines in [CRITICAL LINES]
Use plain language in [BEAT NOTES] – don't worry about being "technical"
One scene per message works best

❌ DON'T:

Don't ask this GPT to generate multiple different scenes at once
Don't add extra sections (like [SUBTITLES]) – it will ignore them anyway

Why we built S.O.R.P.H.I.

Creating high-quality video with AI shouldn't require a degree in prompt engineering. We noticed that while models like Sora are incredibly powerful, they are also sensitive. A slight change in wording can mean the difference between a cinematic masterpiece and a hallucinated mess.

S.O.R.P.H.I. (Sora (2) Output Rendering for Precise High-fidelity Instructions) was designed to bridge that gap. It acts as a translator between your creative vision and the model's technical requirements.

By enforcing a structured input format and converting it into a rigorously tested prompt architecture, SORPHI ensures consistency, continuity, and adherence to your specific direction—things that raw prompting often struggles with.

Frequently Asked Questions

Is SORPHI free to use?

Yes, the custom GPT itself is free to use for anyone with access to ChatGPT. However, to use the generated prompts, you will need access to OpenAI's Sora or a similar video generation model.

Do I need ChatGPT Plus?

Currently, OpenAI requires a ChatGPT Plus, Team, or Enterprise subscription to access custom GPTs like SORPHI.

Can I use the prompts for other video AIs?

While SORPHI is tuned specifically for Sora's understanding of physics, camera movement, and continuity, the detailed shot lists it generates can often be adapted for other high-end video models like Runway Gen-3 or Luma Dream Machine with minor tweaks.

How do I report a bug?

If you encounter issues with the GPT or have suggestions for improvement, please use our Contact page.