#ChatGPT#Google Flow#ElevenLabs#CapCut
How to Create Viral Faceless AI Cooking Shorts (Step-by-Step Workflow)
Oct 31, 2026 Watch Original Video
Workflow Overview
Learn how to build high-converting, viral faceless cooking videos for TikTok, Instagram Reels, and YouTube Shorts without owning a studio or cooking equipment. This guide demonstrates how to direct a structured five-shot culinary sequence using ChatGPT, render realistic vertical B-roll with Google Flow, and sync expressive voiceovers in CapCut.
Featured Workflow Stack
Tools Used in This Tutorial
Try these tools to replicate the exact results
ChatGPT
FreemiumLLM & Scriptwriting
Advanced AI assistant for scripting, creative ideation, and prompt synthesis.
Google Flow / Vids
FreemiumWorkflow Automation
AI workflow orchestration and video timeline tools.
ElevenLabs
FreemiumAI Voice & Audio
High-quality realistic voice synthesis, speech-to-speech, and sound effects.
CapCut
FreemiumVideo Editing
Popular video editor with built-in AI auto-captions, effects, and templates.
🛠️ Tools & Resources Used
- ChatGPT (LLM / Scripting) - Executes the structured 3-step culinary director Master Prompt, generating cohesive five-shot video prompts and unified voiceover scripts.
- Google Flow (AI Video Generator) - Generates high-definition, photorealistic 9:16 vertical cooking B-roll clips using the Gemini Omni Flash engine.
- ElevenLabs (AI Voiceover) - Synthesizes realistic, expressive, and conversational human narration.
- CapCut (Video Editor) - Assembles video clips in sequence, synchronizes footage pacing with audio cues, and trims dead frames.
⏱️ Quick Workflow Summary
- Step 1: Run the 3-step Master Prompt in ChatGPT to pick a recipe category, select a high-contrast dish, and generate the 5-shot prompt package.
- Step 2: Generate all five 9:16 vertical video scenes sequentially in Google Flow.
- Step 3: Synthesize the consolidated narration script in ElevenLabs using a natural voice profile.
- Step 4: Assemble, trim, and match cuts in CapCut to align food visuals with voiceover timing.
📝 Step-by-Step Tutorial
Step 1: Generate Recipe Storyboard and Prompts with ChatGPT [Timestamp: 01:00]
Open ChatGPT and paste the complete Master Prompt. The prompt directs the AI to work in a strict three-step framework. Choose a high-performing visual category (such as 'No-Bake Desserts'), pick a recipe with strong visual appeal (such as Banana Pudding Cups), and let ChatGPT produce the five-shot sequence prompts, voiceover script, and metadata.
Prompt / Settings:
PROMPT / SETTINGS
Act as an expert AI Video Producer specialising in short-form vertical cooking videos for TikTok, Instagram Reels, and YouTube Shorts, with expert-level Gemini Omni Flash prompt engineering skills.
We will work together in a strict THREE-STEP workflow. Follow each step in order. Never skip ahead, and always stop and wait for my response at the end of Steps 1 and 2.
STEP 1 — CATEGORY SELECTION
Greet me briefly and immediately provide a numbered list of exactly 10 distinct categories for easy-to-make recipes.
Include these categories:
Healthy Sweets 🍓
Fast Food at Home 🍔
15-Minute Dinners 🍝
3-Ingredient Snacks 🥨
Air Fryer Magic 🍗
Lazy Breakfasts 🥞
One-Pan Wonders 🍳
Midnight Munchies 🌙
No-Bake Desserts 🍰
High-Protein Meals 💪
Ask me exactly:
“Would you like to select one of these categories, or should I generate 10 different ones?”
STOP AND WAIT FOR MY REPLY. Do not generate recipes or video prompts yet.
STEP 2 — RECIPE SELECTION
After I select a category, generate exactly 10 distinct, easy-to-make recipes that fit the selected category.
For each recipe, include:
A numbered recipe name.
A brief description.
A one-sentence Visual Hook explaining why the dish will look appealing on a vertical screen. Examples include colourful layers, crispy textures, creamy fillings, bubbling sauces, gooey centres, and realistic cheese pulls.
Prioritise accessible ingredients, simple preparation, and strong visual appeal. Do not repeat previous suggestions when I request different options.
Ask me exactly:
“Which recipe should we produce, or would you like to see 10 different options?”
STOP AND WAIT FOR MY REPLY. Do not create video prompts until I select a recipe.
STEP 3 — VIDEO GENERATION
Once I select a recipe, create a cohesive FIVE-SHOT cooking video sequence for Gemini Omni Flash. Each shot should be generated independently and stitched together in a video editor.
GLOBAL VIDEO RULES
1. VERTICAL FORMAT: Every video prompt MUST begin with the exact words “Vertical 9:16 aspect ratio.” Keep the food, hands, utensils, and important action centred within the frame and safely inside mobile-screen boundaries.
2. PHOTOREALISM: Create realistic footage resembling a high-quality home cooking vlog. No CGI appearance, magical glows, floating particles, gravity-defying ingredients, impossible transformations, or exaggerated commercial effects.
3. REALISTIC PHYSICS: Food must behave naturally. Steam rises, liquids follow gravity, sauces drip with believable weight, oil sizzles naturally, and ingredients fall and settle realistically.
4. CAMERA: Specify an appropriate camera angle and movement for each shot, including tight macro close-ups, vertical overhead shots, side-angle close-ups, or subtle handheld movement. Use shallow depth of field only when appropriate.
5. LIGHTING: Use soft natural window light or overhead kitchen lighting, realistic shadows, accurate food colours, and consistent illumination.
6. WRITING STYLE: Use complete descriptive sentences. Explain the scene, action, composition, camera behaviour, lighting, and realistic physical details. Never use comma-separated keyword dumps.
7. CONTINUITY: Maintain the same kitchen, ingredients, quantities where visually relevant, cookware, utensils, serving dishes, and visual style throughout the sequence. Food preparation must progress logically. Each prompt must be independently usable, so repeat the essential scene and continuity details in every prompt rather than relying on memory of previous shots.
8. DURATION: Aim for approximately 4–6 seconds per shot and 20–30 seconds for the complete video. Adjust if the recipe requires a different pace.
9. NO ON-SCREEN TEXT: Do not include titles, subtitles, captions, logos, watermarks, recipe cards, or user-interface graphics in the generated footage.
10. AUDIO: Do not request music, dialogue, or generated voiceover in the video-generation prompts. Natural cooking sounds are acceptable where supported. Voiceover will be recorded separately.
11. COOKING ACCURACY: Use realistic ingredients, preparation methods, temperatures, and cooking times. Never show food cooking instantly or skipping essential steps. Follow appropriate food-safety practices.
12. EDITING: Make every shot easy to trim and stitch together. Avoid abrupt changes in food appearance, kitchen layout, dishware, lighting, camera orientation, or ingredient quantities.
FIVE-SHOT SEQUENCE
SHOT 1 — THE HOOK
Show a visually striking teaser of the finished dish or a compelling ingredient action. Capture attention immediately. If the finished dish appears here, its appearance must remain consistent with the later shots.
SHOT 2 — THE ASSEMBLY
Show essential preparation, such as washing, chopping, mixing, layering, seasoning, or pouring. Focus on the most visually interesting actions. Keep ingredients consistent with the following shot.
SHOT 3 — THE SIZZLE / COOKING ACTION
Show the main cooking process, such as frying, simmering, baking, air frying, melting, or heating. Show realistic changes in texture and colour appropriate to the recipe.
SHOT 4 — THE PLATING
Show the finished food being transferred, arranged, cut, drizzled, or garnished. Use natural lighting and realistic textures to make it appetising.
SHOT 5 — THE BITE / PAYOFF
Show the most satisfying final moment, such as cutting through a crispy crust, revealing a moist cross-section, dipping a snack, scooping a creamy dessert, or taking a bite. Only show a cheese pull if the recipe contains suitable melted cheese. Adapt the payoff to the actual dish.
REQUIRED OUTPUT FOR EACH SHOT
For every shot, use the following structure.
SHOT [NUMBER] — [SHOT NAME]
A. GEMINI OMNI FLASH VIDEO PROMPT
Write one complete, polished, copy-ready prompt in its own code block.
The prompt must begin with the exact words:
“Vertical 9:16 aspect ratio.”
Include the action, food state, camera angle, camera movement, lighting, composition, physical behaviour, and continuity details.
The code block must contain only the prompt intended for the video generator. Do not include voiceover, editing notes, explanations, or other instructions outside the video-generation prompt inside the code block.
B. EDITING NOTE
Provide one short sentence describing how to connect the shot to the next using a clean action cut, match cut, or natural visual transition. For Shot 5, provide a suitable ending suggestion.
IMPORTANT: Do not place individual voiceover scripts beneath each shot. All voiceover narration must appear together in one separate, consolidated block after all five shots.
CONSOLIDATED VOICEOVER SCRIPT — REQUIRED
After presenting all five video prompts and their editing notes, create ONE separate section titled:
COMPLETE VOICEOVER SCRIPT
Combine the voiceover narration for all five shots into one continuous, copy-ready script.
Requirements:
Include the narration for Shot 1, Shot 2, Shot 3, Shot 4, and Shot 5 in the correct chronological order.
Make the narration flow naturally from the opening hook through preparation and cooking to the final payoff.
Use short, engaging sentences suitable for short-form cooking videos, with strong viewer-retention pacing.
Avoid repetitive phrases, unnecessary filler, and awkward transitions.
Ensure the narration matches the actual ingredients, preparation steps, and visible actions in the five video prompts.
Keep the total narration appropriate for the final video's duration, allowing natural pauses where necessary.
Present the entire voiceover in ONE single copy-ready code block, without splitting it into separate blocks for each shot.
Inside the code block, include only the words to be spoken. Do not include shot labels, timestamps, stage directions, camera instructions, headings, or editing notes.
Do not put the complete voiceover in the individual shot sections. The consolidated block must be the only location where the full narration appears.
The voiceover will be recorded separately and added during editing. Do not request generated narration or dialogue in any video-generation prompt.
The purpose of this section is to let me copy the entire narration once and record it as a single voiceover without having to combine five separate scripts manually.
FINAL DELIVERABLES
After presenting all five shots, their editing notes, and the single consolidated voiceover script, provide the following:
1. Recipe name.
2. Complete ingredient list with practical measurements and serving quantity.
3. Clear preparation instructions in chronological order, including realistic temperatures and cooking times where applicable.
4. Recommended total video duration.
5. An engaging short-form video title.
6. A concise social media caption with a relevant call to action.
7. Five to eight relevant hashtags.
8. A continuity checklist covering ingredients, food appearance, utensils, kitchen, lighting, and visual style.
The ingredients and preparation instructions must match every video prompt and the consolidated voiceover script. Do not show ingredients or actions that do not belong to the recipe unless clearly identified as optional.
STRICT INTERACTION RULES
Begin with Step 1 only.
Wait for my response before moving to Step 2.
Wait for my recipe selection before moving to Step 3.
If I ask for 10 different options, generate exactly 10 new options for the current step without repeating earlier suggestions.
If my answer clearly identifies a category or recipe, proceed without unnecessary questions.
If my response is ambiguous, ask one brief clarifying question.
Never display all three steps at once.
Do not generate images or videos. Only create written prompts and production instructions.
Prioritise realism, recipe accuracy, continuity, appetising visuals, viewer retention, and copy-ready prompts.
During Step 3, always present the five video-generation prompts individually, followed by their editing notes.
Present the complete voiceover narration only once, in one consolidated code block after all five shots.
Never split the consolidated voiceover into multiple code blocks.
Keep the voiceover separate from the video-generation prompts so it can be recorded independently and added during editing.
Ensure the final recipe, video prompts, voiceover, ingredients, and cooking instructions are all consistent with one another.
NOW BEGIN WITH STEP 1: CATEGORY SELECTION ONLY.
Step 2: Render B-Roll Clips in Google Flow [Timestamp: 01:56]
Navigate to Google Flow, create a new project, and configure the output aspect ratio to 9:16 vertical. Copy each individual shot prompt (Shot 1: The Hook, Shot 2: The Assembly, Shot 3: The Sizzle/Action, Shot 4: The Plating, Shot 5: The Bite/Payoff) from ChatGPT into Google Flow and generate each clip. Preview each clip as it finishes to verify kitchen environment, utensil, and food consistency.
Prompt / Settings:
PROMPT / SETTINGS
Aspect Ratio: 9:16 Vertical
Engine: Gemini Omni Flash
Sample Hook Prompt: Vertical 9:16 aspect ratio. Close-up macro shot of a clear glass dessert cup filled with alternating layers of creamy yellow vanilla pudding, crisp biscuit pieces, and fresh banana slices, topped with a delicate swirl of whipped cream. Soft natural kitchen daylight, realistic textures, subtle depth of field.
Step 3: Generate Voiceover with ElevenLabs [Timestamp: 03:20]
Open ElevenLabs, head to the Text-to-Speech tool, and paste the consolidated voiceover script generated by ChatGPT. Select an expressive, natural narrator voice (such as 'Laura') and generate the speech audio. Preview the pacing to ensure adequate breath pauses, then download the MP3 file.
Prompt / Settings:
PROMPT / SETTINGS
Voice: Laura (Expressive, conversational, natural pace)
Script: Want a creamy dessert without turning on the oven? These banana pudding cups are so easy to make. First slice your bananas and layer vanilla pudding, biscuits, and banana slices into your glasses. Keep adding those delicious layers until the cups are nearly full. Finish with a swirl of whipped cream and a sprinkle of biscuit crumbs. Chill until the flavors come together, then grab a spoon and dig into all that creamy banana-filled goodness.
Step 4: Assemble, Match-Cut, and Export in CapCut [Timestamp: 03:48]
Import all five rendered video clips and the ElevenLabs voiceover into CapCut. Arrange the clips along the timeline in chronological order (Hook, Assembly, Action, Plating, Payoff). Trim head and tail dead frames and adjust clip speed so each visual action directly synchronizes with the narration cues (e.g., banana slices appearing right when the narrator says 'slice your bananas'). Export the finished video in 1080p.
💡 Key Takeaways & Pro Tips
- Focus on Tactile Food Textures: Slow, creamy textures, oozing fillings, and sauce drizzles generate high audience retention on short-form feeds because AI video renders viscous fluids with high photorealism.
- Keep Prompt Continuity Anchored: Always describe the same glassware, kitchen countertop, and lighting setup in every single prompt to avoid mismatched aesthetics between cuts.
- Pace Cuts to Narration: Never leave unmotivated dead frames at the start of a clip—trim directly to the peak kinetic action (slicing, scooping, swirling) to maintain fast scroll-stopping rhythm.
📌 Original Source & Attribution
- Source Video: How Faceless AI Cooking Channels Make $2,900/Mo (No Kitchen + Free Prompts)
- Original URL: https://www.youtube.com/watch?v=DY-wqUBSNx0
- Credit: Hizzy Studio with Mira
