An account posting five times a week needs two hundred and fifty images a year. Producing them one at a time is both slow and inconsistent.
Batch production solves both at once: you produce a week's images in one session, and because they share a style block they hold together.
The workflow below produces a week of posts in roughly two hours.
Step 0: The one-time job — a style block
Before anything else, turn your brand's visual language into a text block. It will be pasted verbatim at the end of every prompt, and it is what creates consistency.
Three things belong in it: palette colours named explicitly ("muted terracotta, cream, deep olive"), light character ("soft natural window light") and composition preference ("generous negative space, minimal composition").
A stronger method is style reference: pass a reference image with `--sref` in Midjourney, or create a saved style in Leonardo and Recraft. That binds markedly tighter than repeating prompts.
Step 1: List the week's subjects first
Decide the five subjects in writing before generating anything. Thinking about subjects while generating is slow and leads to arbitrary results.
Write a one-sentence visual idea for each. Not "new product announcement" but "product on a wooden counter, morning light, space at the top for text".
Generate all five in one session with the same style block and the same settings. This is the most practical guarantee of consistency.
Step 2: Generate at the right size
An image generated at the wrong ratio gets cropped by the platform, and the most important part of the composition can be lost.
Instagram post 4:5, stories and Reels 9:16, LinkedIn post 1.91:1, YouTube thumbnail 16:9, Pinterest 2:3.
If one image will run on two platforms, generate it twice at two ratios. Changing only the ratio parameter takes seconds and beats cropping.
| Placement | Ratio | Pixels |
|---|---|---|
| Instagram post | 4:5 | 1080 × 1350 |
| Instagram story / Reels | 9:16 | 1080 × 1920 |
| LinkedIn post | 1.91:1 | 1200 × 627 |
| YouTube thumbnail | 16:9 | 1280 × 720 |
| Pinterest pin | 2:3 | 1000 × 1500 |
Step 3: Plan the text area at generation time
If text will sit on the image, reserve that area while generating. Adding "generous empty space in the upper third for text" guarantees it.
Do not have AI render the copy. That gives flawed typography and means a wording change forces a regeneration. Add text in Canva, Figma or any design tool.
The exception is one- or two-word headlines. Ideogram renders those reliably, and it is worth using when you want typography integrated into the image.
Step 4: Prepare the files
The PNGs these tools hand you are large. Platforms compress them anyway, but compressing yourself first gives a better result — platform compression degrades quality more.
Save as WebP or high-quality JPEG and resize to the platform's recommended dimensions.
Name the files by subject and date and keep them in one folder. Six months later you will want to know which prompt produced a given image; writing the prompt into the filename or a text file beside it solves that.
Frequently asked questions
Why don't my posts look consistent?
Because the style description changes with each prompt. Turn your brand colours and visual character into a fixed text block appended to every prompt, and change only the subject. A stronger fix is Midjourney's --sref or Leonardo's saved styles.
Which tool is most practical for social?
If the image becomes part of a design, Canva is fastest because generation and layout live in one place. If the image itself is the deliverable and quality matters, Midjourney or Flux. To stay free, Leonardo AI's daily allowance covers a week's needs.
Should I render the text into the image?
No. Leave empty space and add the copy afterwards in a design tool; that gives error-free typography and means wording changes do not force a regeneration. The exception is one- or two-word headlines, where Ideogram is reliable.