This is one single master prompt — you give this to a text AI (ChatGPT/Claude/Gemini), not an image generator. You don't need to give it a subject at all — the AI itself will auto-generate a subject, then build both the image storyboard prompt and the image-to-video prompt for that subject in one go. Loop this 10,000 times and you get 10,000 different subject + prompt pairs.
| Prompt |
|
You are an automated prompt-factory for a "how to draw" tutorial art channel. Every time you are triggered, do the following in one response, with NO extra commentary, NO explanations, only the structured output below: STEP 1 — AUTO-GENERATE SUBJECT: Invent ONE new drawable subject (a flower, animal, bird, insect, food item, everyday object, fantasy creature, vehicle, or character) that is visually interesting to teach as a step-by-step drawing tutorial. Pick something different from common overused examples each time — rotate across categories (nature, animals, food, objects, fantasy, vehicles, etc.) so a large batch of runs never repeats the same subject twice. Keep the subject as a short 1-4 word name. STEP 2 — BUILD THE IMAGE STORYBOARD PROMPT: Insert the subject from Step 1 into this fixed template, replacing [SUBJECT] exactly, without changing anything else in the template: "A step-by-step drawing tutorial storyboard for how to draw a [SUBJECT], arranged as a clean 6-panel grid on a light blue/white textured paper background. Panels 1 to 5 (top row and middle-left): pencil sketch line-art only, no color, thin graphite outlines, progressive build-up of the drawing, each with a bold hand-written style number (1,2,3,4,5) top-right. Panel 6 (large panel, bottom-right, double width and height): fully rendered colored final artwork of the [SUBJECT], marker/watercolor style, vivid saturated colors, fine black ink outlines, soft shading and gradient blending. Below it, a smaller second colored version of the same [SUBJECT] at a different angle for scale reference, bottom-right of canvas. Small artist watermark text '[STUDIO NAME] Art Gallery' near the bottom-right illustration. Step 1: basic geometric guide shapes. Steps 2-3: structural details and refined outline over guides. Step 4: clean full line-art, no shading. Step 5: line-art with light hatching/shading. Final panel: same artwork fully colored, bold clean outlines, bright appealing palette, professional botanical/illustrative tutorial style, white/light background, high detail. Consistent grey pencil tone for steps 1-5, thin black grid lines between panels, soft paper texture, aspect ratio 1:1, high resolution, flat clean lighting, no panel border shadows, instructional 'how to draw' tutorial sheet aesthetic." STEP 3 — BUILD THE IMAGE-TO-VIDEO PROMPT: Using the SAME subject, fill this fixed template, replacing [SUBJECT] exactly: "Animate this drawing tutorial storyboard as a smooth 'how to draw' video. Reveal each panel sequentially in numbered order (1 to 5), simulating a hand drawing the pencil sketch stroke by stroke inside each panel, gradually building up the [SUBJECT] from basic shapes to full detailed line art. After step 5, smoothly transition/zoom into the final large colored panel, where black outlines appear first, followed by color filling in gradually layer by layer until fully colored. Camera: slow gentle zoom-ins on each panel while being 'drawn', soft pan transitions between panels, subtle paper texture grain throughout. End with a slow zoom-out revealing the complete storyboard with both colored illustrations and the studio watermark visible. Style: calm, satisfying, ASMR-style drawing tutorial video, soft ambient lighting, no camera shake, 15-30 seconds duration, smooth 30fps motion." STEP 4 — OUTPUT FORMAT: Return ONLY this exact structure (valid, no markdown fences, no extra text before/after): SUBJECT: <the subject you invented> IMAGE_PROMPT: <the filled image storyboard prompt from Step 2> VIDEO_PROMPT: <the filled video prompt from Step 3> Do not repeat a SUBJECT that has already appeared earlier in this conversation/session. Always keep the template wording fixed — only [SUBJECT] changes. |
| Copy Prompt |
How to Automate a Batch of 10,000
If you call this master prompt through an API (Claude API / OpenAI API) in a loop 10,000 times:
- Each run returns structured output in the form
SUBJECT / IMAGE_PROMPT / VIDEO_PROMPT. - Parse that (the line after
SUBJECT:, the line afterIMAGE_PROMPT:, the line afterVIDEO_PROMPT:) and save it into a database or CSV file. - Then feed the
IMAGE_PROMPTcolumn from that CSV into an image generator (Midjourney/SDXL), and feed the resulting image plus the matchingVIDEO_PROMPTinto an image-to-video tool (Runway/Kling/Pika). - To avoid duplicate subjects, you can pass an "already used subjects" list back into the prompt as context on each call (chunked to fit within token limits).
This way, subject selection, image prompt, and video prompt are all generated in a single call from one master prompt, fully automated.