How to Make AI Videos From Text and Images
Short answer: To make an AI video, pick a text-to-video tool such as Google Veo (via Gemini or Flow), OpenAI's Sora, or Google Vids for slide-style videos. Write a shot-list prompt describing scene, camera, motion and audio, generate a short clip of a few seconds, then join clips and add voice in an editor.
AI video tools now produce clips that look convincing at a glance. They also burn through credits fast, and most beginners throw away far more clips than they keep. The skill is not finding a tool; it is writing prompts that land a usable clip in two or three tries.
This guide covers the shot-list method we use, how to work out what a usable clip costs you, and the giveaways that still make AI video look fake.
Which AI video tools can you use?
Tools fall into three groups: generators that create footage from a prompt, editors that build videos from slides, stock and voice, and avatar tools. Availability and limits change monthly, so treat this as a map, not a price list.
| Tool type | Examples | Best for | Typical access (as of November 2026) |
|---|---|---|---|
| Text/image to video | Google Veo (in Gemini app and Flow), OpenAI Sora, Kling, Runway | Short cinematic clips, product shots, b-roll | Limited free trials or daily quotas; real use needs a paid plan or credits |
| Template video editors | Google Vids, Canva | Explainers, training videos, slide-to-video | Free tiers exist; AI features vary by plan |
| Avatar and voice | Avatar presenters, AI voice tools | Talking-head explainers, voiceovers | Usually credit or minute based |
Sora vs Veo is a common comparison. Both make short clips with synchronised sound in their recent versions; the practical difference for most users is access. Veo comes bundled in Google AI plans many Indians already have through offers, while Sora's availability and app access have varied by country. Check what works on your account today rather than trusting a review from three months ago.
The shot-list prompt method
Write each clip as one shot from a film crew's shot list: scene, camera, motion, audio. Video models handle one clear shot far better than a story crammed into one prompt.
- Scene: subject, place, time of day, lighting, style. 'A steel tiffin box on a wooden table by a window in a Delhi kitchen, soft morning light, realistic.'
- Camera: shot size and angle. 'Close-up, eye level, shallow depth of field.'
- Motion: one movement for the subject and one for the camera. 'Steam rises as the lid lifts; slow push-in.'
- Audio: what we hear, if the tool supports sound. 'Quiet kitchen ambience, a lid clink, no music, no speech.'
- Limits: length, aspect ratio, what to avoid. '8 seconds, vertical 9:16, no text, no people.'
Plan the full video as 4–6 such shots, generate each separately, then join them. For a 30-second reel, that is usually five or six clips plus a voiceover. The same structure applies to still images, which is why practising with image prompts first is cheaper: it costs fewer credits to learn what 'soft morning light' produces. For the general prompting principles behind this, see our checklist for writing AI prompts.
A worked example: before and after
Say you are making a reel for a Jaipur block-print brand.
Before: 'Make a beautiful video of block printing.' Results are typically generic: random hands, a fabric that changes pattern mid-clip, and a camera that drifts for no reason.
After: 'Close-up of a craftsman's hands pressing a carved wooden block onto white cotton fabric on a long printing table, indigo dye, natural daylight from the left, realistic documentary style. Camera: static, top-down at a slight angle. Motion: the block lifts slowly to reveal the printed motif. Audio: a soft thud and cloth rustle, no music. 8 seconds, 9:16.'
- One action (press and lift) gives the model a clear beginning and end.
- A static camera reduces the wobble and morphing that give AI video away.
- Naming the dye and light direction keeps colour consistent across clips you will join later.
- Check the result frame by frame for hands. Fingers merging into the block is the most common flaw in this kind of shot.
What does a usable AI clip actually cost?
Work out cost per usable clip, not cost per generation. Most tools charge per generation, and you will reject many of them.
Here is the calculation with illustrative numbers. Say a plan costs you around ₹2,000 a month and allows roughly 60 short clip generations. That is about ₹33 per generation. If 1 in 3 is usable, each usable clip costs about ₹100. A 30-second reel needing 6 usable clips then costs around ₹600 in generations, before your time.
| Assumption | Value (illustrative) | Your number |
|---|---|---|
| Monthly plan cost | ≈ ₹2,000 | Check current pricing |
| Generations included | ≈ 60 | From the plan's credit table |
| Cost per generation | ≈ ₹33 | Plan cost ÷ generations |
| Keep rate | 1 in 3 | Track your own for a week |
| Cost per usable clip | ≈ ₹100 | Cost per generation × tries |
These are not quoted prices; plans, credits per clip and quality modes differ by tool and change often. The point is the method. Better prompts raise your keep rate, and that cuts cost faster than switching tools. Faster or lower-quality modes cost fewer credits, so draft in the cheap mode and use the high-quality mode only for the final take.
How to make a video with Google Vids or a template editor
If you need a training video, a product explainer or a slide-based video rather than cinematic footage, a template editor is faster and cheaper than a generator.
- Open Google Vids from Google Drive or Workspace (or a similar editor such as Canva).
- Describe the video or upload a document, such as a product sheet or SOP; the tool drafts a storyboard.
- Edit the scenes and script. Remove any claims the AI added that you cannot back up.
- Add stock clips, your own phone footage, or a few AI-generated shots.
- Add a voiceover: record your own or use an AI voice. For Hindi narration, compare options in our guide to Hindi AI voice generators.
- Export and watch the whole video on a phone before posting.
What still looks fake in AI video?
Short clips can fool people, but some flaws remain common enough that viewers notice them.
- Hands and fingers, especially when holding tools or touching faces.
- Text on signs, packaging and screens: often garbled or changing between frames.
- Physics: liquid pouring the wrong way, fabric passing through objects, food that changes shape.
- Continuity between clips: a character's clothes, face or the product changing from shot to shot.
- Lip sync on Hindi dialogue, which is generally weaker than English.
- Indian specifics: real brand packaging, accurate local signage and crowded street scenes often come out wrong or generic.
A quick quality check before you post: watch each clip three times, once at normal speed, once paused frame by frame at the start and end (where morphing hides), and once with sound off to see if the motion still makes sense. Delete anything that fails one of the three.
Practical rule from client-style work: never generate your actual product. Shoot the product on a phone and use AI only for backgrounds, b-roll and mood shots. That avoids the product's label or shape changing, which is both a quality problem and a potential consumer-complaint problem.
Rules and labels you should know
Platforms increasingly require you to label realistic AI-generated content, and some tools add visible or invisible watermarks. Do not make videos of real people without consent, do not imitate public figures, and do not present AI footage as a real event. If the video is an ad, ad platforms have their own disclosure rules, which are a separate topic from this general how-to.
For still visuals to go with your videos, see our comparison of free AI image generators. The shot-list method, voice tools and editing are practised hands-on in our 8-week applied AI course.
Frequently asked questions
Can I make AI videos for free?
Yes, within limits. As of November 2026 several tools offer free trials, small daily quotas or bundled access through Google AI offers in India, but serious use usually needs a paid plan or credits. Check each tool's current free allowance.
How long can an AI-generated video clip be?
Most text-to-video tools generate clips of a few seconds to a little over ten seconds at a time. Longer videos are made by generating several clips and joining them, or by extending a clip where the tool allows it.
Which is better, Sora or Veo?
Both produce strong short clips, and quality leadership changes with each update. For most Indian users the deciding factor is access and cost, so test both on the same shot-list prompt if you can.
Can AI make a video with Hindi voiceover?
Yes. You can generate the visuals with a video tool and add a Hindi voiceover from an AI voice tool or your own recording. Check pronunciation of names and brand words before publishing.
Can I use AI videos on Instagram and YouTube?
Yes, but follow each platform's rules on labelling realistic AI content, and make sure you have rights to any music, faces or brands that appear.
Keep reading in AI tools
- How to Write AI Prompts That Work First Time Start here
- Is It Safe to Upload Documents to ChatGPT? Privacy Settings and What Not to Share
- How to Use Deep Research in ChatGPT and Gemini: Brief It, Then Audit It
- How to Fact-Check AI Answers Before You Use Them
- Free AI subscriptions in India and how to claim them
Courses: Applied AI Professional · AI-Powered Digital Marketing · All guides in AI tools