Quick answer
Split the work into planning, images, animation, editing and review. Use ChatGPT for the brief, script, shot list and prompts. Use Midjourney to make still frames and animate them into short clips. Then assemble clips, narration, music and captions in a video editor, and check rights and facts before export.
You can create an AI video with ChatGPT and Midjourney by splitting the work into planning, image creation, animation, editing, and review. ChatGPT is useful for the concept, script, shot list, narration, captions, and prompt variations. Midjourney can create the still images that become your starting frames, and its current video workflow can animate an image into a short sequence. You still need a video editor for timing, sound, captions, and export.
One important update: older tutorials often say to generate the finished video inside ChatGPT with Sora. OpenAI’s Sora web and app experiences were discontinued on April 26, 2026, so treat ChatGPT as the planning and writing layer unless you have a separate connected video tool. The workflow below is designed to remain useful when the rendering tool changes.
The workflow at a glance
- ChatGPT: develop the idea, audience, script, shot list, and image prompts.
- Midjourney: create consistent stills and animate selected images into short clips.
- Editor: arrange clips, add narration, music, captions, transitions, and branding.
- Review: check continuity, rights, factual claims, accessibility, and platform specifications.
Step 1: Define the video before generating anything
Start with a brief instead of asking for “a cool AI video.” Tell ChatGPT the topic, audience, goal, platform, duration, aspect ratio, tone, visual style, narration language, and any subjects or claims that must be avoided. A clear brief prevents attractive but unusable images.
Try this planning prompt:
Create a production brief for a [30-second] [9:16] video about [topic] for [audience] on [platform]. Return: one-sentence concept, hook, five-shot outline, narration under [word count] words, on-screen text, visual continuity rules, sound notes, and a final checklist. Keep every claim verifiable and mark anything that needs fact-checking.
Step 2: Turn the idea into a shot list
Ask for one row per shot. Include duration, visible action, camera position, lighting, subject placement, transition, narration, and the exact image prompt. Keep the same character description, wardrobe, color palette, lens language, and environment across shots. Consistency is easier when you generate fewer, stronger frames than dozens of unrelated images.
Use this prompt to refine it:
Convert this brief into a five-shot table. For each shot provide: timecode, purpose, action, composition, camera movement to simulate in editing, lighting, continuity details, narration, on-screen caption, and a Midjourney image prompt. Repeat the locked character and style details in every prompt.
Step 3: Create a visual bible
Before generating images, ask ChatGPT for a short visual bible. Lock the main subject’s age range, clothing, colors, hairstyle, props, location, time of day, art direction, and camera style. Add a negative list such as extra fingers, illegible text, duplicated objects, inconsistent logos, and random costume changes. For a product or brand video, upload or use approved reference assets when the tool supports them.
Step 4: Generate the starting frames in Midjourney
Use the image prompts from the shot list in Midjourney. Choose the aspect ratio that matches the final destination: 9:16 for vertical short-form video, 16:9 for landscape video, or 1:1 for square social posts. Generate several variations, select frames with clean composition, and keep a record of the prompt and settings for every approved image.
A useful image prompt structure is:
[subject] doing [specific action], [location and time], [composition], [lighting], [lens or camera feel], [color palette], [visual style], clean background, no readable text, no watermark, consistent character design –ar 9:16
Do not rely on generated text for titles, labels, or disclaimers. Add those in the editor so they remain legible and editable.
Step 5: Animate selected images
Midjourney’s current Video feature starts with an image and can generate a moving sequence from it. Use a restrained motion prompt: describe the subject’s movement, camera movement, and pace, then specify what must remain still. For example: slow push-in, leaves moving gently in the wind, subject maintains the same face and clothing, stable background, natural motion.
- Prefer subtle motion when the image contains faces, hands, products, or text.
- Use the same starting frame again when you need a loop or a matching transition.
- Reject clips with warped objects, flicker, identity changes, or impossible physics.
- Keep the original stills so you can create a slideshow or alternate edit if animation fails.
Step 6: Assemble the video in an editor
Import the selected clips and stills into a video editor such as Premiere Pro, DaVinci Resolve, CapCut, or another tool you already use. Build the rough cut first, then add narration, music, sound effects, captions, a cover frame, and a final call to action. Use slow zooms or pans on still images rather than forcing every frame into an unstable animation.
Ask ChatGPT to review the edit plan with this prompt:
Review this shot list as a video editor. Identify pacing problems, repeated visuals, unclear narration, caption-length issues, continuity errors, accessibility risks, and claims that need fact-checking. Return a prioritized revision checklist for a [platform] video.
Step 7: Export and quality-check
- Watch the final export on a phone and desktop, with sound on and off.
- Check that captions are readable, synchronized, and not hidden by platform controls.
- Confirm that music, voices, logos, references, and likenesses are used with appropriate permission.
- Disclose synthetic or materially altered media when the platform or audience reasonably needs that context.
- Export a high-quality master and a platform-specific copy; keep the project file and source assets.
Common mistakes to avoid
- Following an outdated Sora tutorial as if Sora were still available.
- Changing the character description or aspect ratio from shot to shot.
- Generating a full minute before testing one short clip.
- Letting an image model create important text, prices, dates, or product specifications.
- Publishing AI footage without checking rights, factual accuracy, or visual artifacts.
Frequently asked questions
Can ChatGPT generate the finished video by itself?
ChatGPT can help plan and write the video, but the rendering step depends on the tools currently available in your account and workflow. OpenAI’s Sora web and app experiences were discontinued in 2026, so do not promise native ChatGPT video generation without checking the current product.
Can Midjourney turn an image into a video?
Yes. Midjourney’s current Video documentation describes using an image as a starting frame, optionally adding a text prompt, and generating a short moving sequence. Features and limits can change, so check the official documentation before production.
Is this workflow suitable for commercial content?
It can be, but review each tool’s current terms, plan permissions, music and voice licenses, model-release needs, trademarks, and platform disclosure rules. Keep records of your prompts, source assets, and approvals.
Final takeaway
The most reliable way to create an AI video with ChatGPT and Midjourney is to separate creative planning from rendering. Let ChatGPT create a precise brief, shot list, script, and review checklist; use Midjourney for controlled starting frames and short motion; then finish the story in an editor. Test a small sequence first, preserve your source files, and verify current tool availability before promising a particular feature.
For more visual prompt ideas, explore our Midjourney prompt collection and our guide to writing better GPT-4 prompts.
For a broader library of reusable AI workflows, explore the Complete Playbook Bundle and use our GPT-4 prompting framework when planning scripts, shot lists, and reviews.
2 thoughts on “How to Create an AI Video with ChatGPT and Midjourney”