ChatGPT math explainer

Take a photo of any equation, drop it into ChatGPT, and get back a fully narrated video that walks through the solution one step at a time. Here's the exact setup, start to finish.

ChatGPT does the math. Higgsfield turns it into video.

ChatGPT is a genuinely good tutor — it can read your problem and solve it correctly, showing every step. But on its own it can only talk. It can't draw, animate, or render a video of what it's explaining. That's the missing piece.

The Higgsfield connector bolts a video engine straight onto ChatGPT. Once it's connected, the same chat that solves your problem can also produce a clean, narrated explainer video of that exact solution — without leaving the conversation.

The order matters, and it's why this stays accurate: ChatGPT solves the math in text first (that's where the real working lives, and where you confirm the answer is right), then hands the finished steps to Higgsfield to visualize. The narration carries the correct reasoning; the video just makes it easy to follow.

This is the part people miss: ChatGPT can't make this kind of video by itself. The connector is what gives it access to Higgsfield's video models inside the chat — so setting that up is genuinely step one.

The usual way

Search YouTube for a video that's kind of like your problem, scrub through 15 minutes, and hope the numbers line up with yours.

With Higgsfield in ChatGPT

Photograph your exact problem and get a short, narrated video solving that one — on demand, in a couple of minutes.

Download the Whiteboard Skill

This is the skill that runs everything — download it first. It makes ChatGPT solve your problem, verify the answer, then generate one continuous top-down whiteboard video with synced narration. Grab it here, then do the quick setup below; you hand it to ChatGPT in Step 4.

---
name: solve-math-whiteboard-video
description: Turn an uploaded photograph or screenshot of a math problem into a verified step-by-step solution, a clean handwritten whiteboard end-frame image, and a concise narrated instructional video. Use when the user asks to solve math from an image and demonstrate the work visually, especially for whiteboard lessons, handwritten solution videos, math explainers, or Higgsfield video generation. The video must use one continuous top-down shot with no cuts, scenes, angle changes, or cinematic treatment.
---

# Solve Math Whiteboard Video

Create a complete math mini-lesson from one uploaded equation image. Prioritize mathematical correctness, legibility, and instructional synchronization over visual entertainment.

## Required workflow

### 1. Read and transcribe

1. Inspect the uploaded image.
2. Transcribe every problem exactly, including restrictions, radicals, fractions, exponents, signs, and rounding directions.
3. State any ambiguous character before proceeding. Do not silently guess ambiguous math.

### 2. Solve and verify

1. Solve each problem independently.
2. Write one meaningful algebra operation per line.
3. Preserve domain restrictions throughout the work.
4. Verify each final answer by substitution or an equivalent independent check.
5. Apply requested rounding only after finding the exact solution.

Do not generate media until every answer passes verification.

### 3. Design the end frame

Create a clean whiteboard image containing the original problem, every solution step in order, exact solutions before rounded solutions, and a clearly circled final answer. Use black handwritten dry-erase writing, a plain white board, ample spacing, and no unrelated decoration.

Treat this completed board as the authoritative final frame and accuracy reference for the video. Inspect the generated image and reject it if any mathematical symbol, number, line order, restriction, or answer is wrong.

### 4. Write the lesson script

- Use a direct teacher-like tone.
- Explain each operation as the hand writes that line.
- Avoid hooks, jokes, cinematic language, dramatic pauses, or motivational filler.
- Spell mathematical notation naturally for speech.
- Divide narration into fixed time windows only when the video tool requires blocks.
- Keep one logical section within each window and prevent narration from bleeding into the next section.

### 5. Generate narration

Generate narration before final assembly so measured duration can guide pacing.

1. Use one locked narrator voice throughout.
2. Measure speech duration rather than relying only on word count.
3. Rewrite wording if a take overruns its visual window.
4. Do not time-stretch, pitch-shift, or artificially accelerate narration.
5. Preserve completed takes when retrying another section.

If the video workflow requires fixed ten-second blocks, lock those windows first, then generate narration to fit each window.

### 6. Generate the whiteboard video

Read references/video-prompt.md before generating video.

#### Hard visual rules

- Use exactly one continuous shot for the entire video.
- Use a fixed top-down view of the same whiteboard.
- Do not use cuts, inserts, scene changes, shot lists, angle changes, or cutaways.
- Show only the board, writing hand, and plain black marker.
- Have the hand write each verified solution line in order.
- End on the approved completed whiteboard image.
- Allow only a subtle continuous slow zoom-in when movement is necessary.
- Never use cinematic lighting, dramatic camera motion, animated graphics, captions, or decorative effects unless explicitly requested.

If a system mandates multiple duration blocks, every block must continue the same uninterrupted composition. Prompt later blocks as seamless continuations of the same top-down take. Never introduce cuts within or between blocks.

### 7. Assemble and verify

1. Assemble one final video file.
2. Keep narration synchronized with the line being written.
3. Confirm duration, aspect ratio, and narration presence.
4. Visually confirm there are no cuts or angle changes.
5. Confirm the final frame matches the approved reference.
6. Confirm every displayed equation is correct and legible.

Do not deliver a video that fails any check. Report the precise blocker instead of substituting a cinematic or multi-shot version.

## Default choices

- format: 16:9 landscape;
- duration: 30 seconds;
- subtitles: off;
- treatment: plain instructional whiteboard;
- camera: continuous top-down shot;
- motion: subtle slow zoom-in only;
- narrator: reuse a known selected voice; otherwise use the available voice picker.

## Output order

1. Exact transcription.
2. Verified solution and check.
3. Completed whiteboard image.
4. Narration script.
5. Single-shot narrated video.

~/.claude/skills/solve-math-whiteboard-video/SKILL.md

No connector, no video. The prompt solves the math either way, but it can only render the video once Higgsfield is connected.

• Upload a calculus problem and ask for a video that names each rule as it's used.

• Photograph a geometry proof and get a narrated walkthrough of every step.

• Drop in an algebra word problem and have it explain how to set up the equation before solving.

• Paste a problem you got wrong and ask the video to point out where people usually slip up.

• Ask for a slower version of a concept you're still shaky on.

• Request a recap scene at the end that summarizes the method in one line.

This is one workflow. If you want the rest of the system — the automations I actually run my content and my business on — that's what I share on @the.agi.guide.

Turn your next math problem into a video tonight

Connect Higgsfield to ChatGPT once, and every problem — homework, test prep, that one concept that never clicked — becomes a short video that actually explains it.