How to Create AI Images and Videos for a Business Free in 2026: Product Shots, Posts, Short Ads (ChatGPT, Gemini, Canva, Jobbit)

The candle maker needs a photo of the new jar on a kitchen table in autumn light, a square version for Instagram, a tall one for a story, a banner for the website and a ten-second clip of the flame for the reel, and she has a phone, a kitchen and no afternoon. A year ago that was a photographer, a designer and £400. Today it is a description. To create AI images and short videos for a business is now a sentence to a model, and the question has moved from "can it" to "which one, what does it cost, and what am I allowed to do with the result".

The free tiers are real but tight. ChatGPT gives a free user two or three images a day; Gemini allows a daily batch with its Nano Banana model; Canva's Magic Media sits inside Canva Pro at £13 a month; Kling hands out 66 free credits a day with a watermark and Runway 125 credits once (as published on each company's pages and recorded in our September 2026 ranking of 40 generators). This guide covers the 12 visuals a small business actually needs, the prompts that produce them, what each tool allows before you pay, the UK rules on advertising and ownership, and the one message that makes, resizes and captions everything on Jobbit, where the Free plan gives 100 credits a day with no card.

Create the images and videos on Jobbit in one message

The Jobbit agent generates images, video, audio and music in its own workspace, with a choice of generation models on paid plans, and hands the files back in the chat. The useful habit is to ask for the whole set at once, because the resizing and the captions are where the time goes, not the generation:

1Make a product image of this candle (photo attached) on a wooden kitchen table in soft autumn morning light, no text, the label readable. Then give me: a 1080 by 1080 version for Instagram, a 1080 by 1920 story with the words "Autumn is back" in a clean serif at the top, a 1920 by 600 website banner with space on the left for a headline, and three alternative captions under 150 characters. Keep the jar exactly as in my photo.

The agent returns the master image, the three sizes, the captions and a note on anything it changed. The next message can be "a ten-second clip of the flame flickering with the jar still, vertical, no sound", "the same scene in winter", or "remove the background and give me a PNG for the shop listing". Sizes for every platform are in How to Resize and Convert Images for Free, and the image resizer page does the batch in one go.

12 visuals a small business needs, and how to get each one

The pattern: images are easy and cheap, text inside images is still unreliable, and video is good at mood and movement and bad at hands, faces, logos and lettering. Plan around those four weaknesses and the free tiers cover a month of posts.

Images: the prompts that work for a small business

Product on white

Attach your own photo and say "keep the product exactly as photographed". Models that edit a reference image (Gemini's Nano Banana, ChatGPT's image model) hold the shape and the label far better than models that generate from text alone, which will invent a plausible but different jar. Ask for a PNG with the background removed if the shop listing needs it, and check the label at full size: letters are the first thing to drift.

Lifestyle scenes

Describe the place, the light and the mood, not the camera settings: "on a wooden kitchen table, soft morning light from the left, a linen cloth, out-of-focus window behind". Say what must not be there ("no people, no text, no other products"). Generate four, pick one, then ask for variations of that one rather than new prompts.

Social graphics with text

Text is where image models still fail, so either ask for a clean image with "space for a headline at the top" and add the words in a design tool or in the resizer step, or use a model that handles lettering (Ideogram is built for it). Keep words to a maximum of five. Check the spelling letter by letter.

Consistency across a set

Give every prompt the same reference image and the same two-sentence style description, and ask the agent to "keep the style of the first image". For a product range, generate one master scene and ask for each product dropped into it, rather than a new scene per product.

Short videos: what 15 seconds can and cannot do

AI video in 2026 does three things well for a small business: slow movement over a still scene (a push-in on the product, steam rising, a flame flickering), ambience for a shop or a service (a sunlit workshop, rain on a cafe window), and text-on-motion when the words are added afterwards. It still does four things badly: hands doing anything precise, real people's faces, legible logos and lettering, and anything longer than about 15 seconds without a cut.

So brief it like a photographer, not a film director: "vertical, ten seconds, a slow push-in on the jar on the table, flame flickering, nothing else moves, no text". Generate two, keep the better, add your words and a voice-over afterwards. Google's Veo (inside Gemini on the paid plans), Kling and Runway are the three most people use; all three give a watermarked or limited free taste and charge from about $10 to $20 a month, and the full comparison with current models is in Best AI Image, Video, Voice and Music Generators in 2026.

What the big names allow for free in 2026

ChatGPT's free plan gives two or three images a day and sells Plus at about £20 a month. Gemini has a daily allowance of Nano Banana images on the free plan, with more on Google AI Plus at £4.50 a month and Veo video clips on AI Pro at $19.99. Canva's Magic Media is inside Canva Pro at £13 a month with a monthly cap of uses. Midjourney has no free tier and starts at $10 a month. Ideogram gives 10 images a day free. Adobe Firefly gives about 25 credits a month free and starts at $9.99, with the commercial indemnity that larger clients ask about. Kling gives 66 watermarked credits a day; Runway gives 125 credits once, then $12 a month on the annual plan. Jobbit's Free plan gives 100 credits a day for the whole agent, images included, with no card, and the paid plans from £7 a month add the choice of generation model.

The honest summary is that the free tiers will make your first month of posts if you are disciplined, and the subscriptions buy volume and better models. The subscription trap is paying three of them for one job, when a single agent with model choice does all of it on one plan.

Rights, honesty and the UK rules

• Advertising. The ASA's basic rule applies to AI images exactly as to photographs: an ad must not mislead. An AI picture of your candle on a table is fine; an AI picture that adds a feature the product does not have, a bigger size, a gift box that is not included, is not.
• Real people. Do not generate images of real staff, customers or public figures without consent; for a face, use a real photo or keep faces out of frame.
• Ownership and licences. Who owns an AI image varies by country and by tool; what matters day to day is the tool's terms. Most paid plans allow commercial use of what you generate; some free tiers do not, and some add a watermark. Read the line about commercial use before the first post.
• Other people's work. Do not prompt with a competitor's photo or a named artist's style for something you will sell. The risk is legal and reputational.
• Say so when it matters. If an image shows what a product could look like rather than what it looks like, label it as a visualisation. Customers forgive a render; they do not forgive a surprise.

How an AI agent helps, and the bigger job it can do

ChatGPT, Claude, Gemini, Microsoft Copilot, Manus and Meta's Muse will all make or caption an image, and the image models inside ChatGPT and Gemini are excellent. Where Jobbit differs is what surrounds the generation:

• A choice of models for the job. Reference-faithful editing for product shots, lettering-friendly models for graphics, video models for clips, chosen per task on paid plans.
• Every size in one go. The master image becomes the Instagram square, the story, the banner, the thumbnail and the under-200 KB web version through the resizer, correctly named.
• Captions, alt text and a posting plan. Written from the image and your product, in your voice, with the alt text your website needs.
• The product page or the shop. The images go straight into a page the agent builds and hosts on your domain, on the Go plan at £7 a month.
• A photographer when you need one. For the hero shot of a product that must be exactly right, Jobbit can bring a UK photographer in from the same place, with payment held in escrow until you have the photos.

Pauline, 66, makes soy candles in a converted shed. Her first message produced the autumn scene, the three sizes and the captions. The second produced the ten-second flame clip. The third built the shop page and the fourth added a seasonal collection with the same scene in winter. The images were the smallest thing the agent made; the shop is what the images were for. How to Make a Logo for Your Business covers the one visual where vector files matter, and AI Product Photo Generator goes deeper on studio-style product shots.

Mistakes that make AI visuals look like AI visuals

• Too much in the prompt. Twelve adjectives produce a muddle. Place, light, mood, what must not be there.
• No reference photo. The product in the picture is not your product.
• Text in the image. Spelling drift and odd letters. Add words afterwards.
• Hands and faces. Keep them out or expect to regenerate.
• A different style every post. Reuse the master scene and the style sentence.
• Posting the full-size file. A 4 MB image slows the page and gets cropped. Resize first; How to Compress Images for a Website has the sizes.
• Ignoring the licence line. Free tiers with no commercial use, watermarks you did not notice.
• Misleading edits. The gift box that is not included, the colour that does not exist.

A one-week plan for a month of visuals

• Day 1: the master scene. One product, one place, one light. Four versions, keep one.
• Day 2: the range. Each product dropped into the master scene; check every label.
• Day 3: the sizes. Square, story, banner, thumbnail, web, through the image resizer page, named by product and size.
• Day 4: the clip. One ten-second push-in, vertical, words added afterwards.
• Day 5: the words. Captions, alt text, a posting calendar for four weeks.
• Day 6: the page. The images on a product page or a one-page shop, on your own domain.
• Day 7: the check. Every image at phone size, every price against your list, every licence line read.

Frequently asked questions

Can I create AI images for my business for free?

Yes, within daily limits: ChatGPT gives two or three images a day free, Gemini a daily allowance, Ideogram 10 a day, and Jobbit 100 credits a day on the Free plan with no card. The limits cover a few posts a week; a month of content for several products usually means a paid plan or an agent that does the generating, resizing and captioning on one plan.

Which AI is best for product photos?

Models that edit a reference image rather than invent from text: Gemini's Nano Banana and ChatGPT's image model hold the product's shape and label best, and the Jobbit agent uses a reference-faithful model when you attach your own photo. Always compare the result with the real product at full size.

Can AI make a short video for my business?

Yes, for mood and movement: a slow push-in on a product, steam, flames, rain on a window, 5 to 15 seconds, vertical for reels and stories. It is still poor at hands, faces, logos and lettering, so add the words afterwards and keep real people out of the frame.

Am I allowed to use AI images in adverts in the UK?

Yes, as long as the ad does not mislead, which is the ASA's standard rule. Do not add features, sizes or extras the product does not have, do not generate real people without consent, and label visualisations as visualisations.

Who owns an AI image I generate?

It depends on the tool's terms and on the law where you are, and the terms are what matter in practice: most paid plans allow commercial use of your outputs, some free tiers do not, and some watermark. Read the commercial-use line before you post.

How do I get consistent images across my products?

Make one master scene and reuse it: attach the same reference image and the same two-sentence style description to every prompt, generate each product into that scene, and ask the agent to keep the style of the first image.

What is the fastest way to get a month of visuals?

One message to an agent that generates, resizes and captions: describe the scene, attach the product photo, list the sizes and ask for the captions. On Jobbit the Free plan gives 100 credits a day with no card, the resizer page does every platform size in one zip, and the same chat builds the page the images are for.