Pika: AI video from photos – how to use it for free in 2026

Section 01What Pika is in 2026 and what changed

Pika launched in 2023 as one of the first AI tools that made video from text and images. People remember it for the funny effects: the person in a photo gets squished, melts, or turns into a cake. Those clips went viral, and many people still think Pika is a meme toy.

Today it's a different story. On September 17, 2026, Pika launched a new platform that brings video, images, and sound together. One account now runs Pika's own models alongside models from other developers: ByteDance, Google, Kling, Alibaba, MiniMax, OpenAI. You pay from one credit balance and choose which model makes each clip.

The old version of the site moved to old.pika.art. It says so right there: Pika has changed, and the old version is available temporarily. If you watched Pika videos from a year ago and are looking for the same buttons, they're gone from the main site. The funny Pikaffects (squish, melt, cake-ify) live on in the Pika mobile app and on the old version of the site.

There are two fresh additions too. On October 16, 2026, Pika showed up inside ChatGPT: you can ask the chat to make a clip, and it calls Pika's tools. On October 29, Pika H3 Flex came out – a model built on MiniMax H3 that takes up to 12 references, makes video up to 15 seconds with sound, and upscales the picture to 4K. You won't need it for your first clips, but if you see the name in the model list, you'll know what it is.

The Pika interface is in English. There are a lot of buttons, but one app is enough to start: Video Studio. Anything you need to write as text, Claude can draft for you – more on that in Section 10.

Section 02Pika for free: what you really get

This is the main question people bring to Pika, so I'll answer it straight. In early 2026 the free plan gave 80 credits a month, enough for a few clips. That's over. On the official pricing page the Free plan says: 0 credits / month · packs only. In other words, zero credits a month, and you can only generate with purchased packs.

What the Free plan itself gives you: an account, access to all the models and apps, up to two videos generating at once, and clips without a watermark. But there's no commercial license on Free: clips from the free plan can't be used in ads or in work for clients.

The only free route: the ChatGPT bonus

The Pika app catalog has an offer: make your first Pika generation inside ChatGPT and get 250 bonus credits. It's a one-time gift, it doesn't renew every month. But it's enough to get to know the tool.

What 250 credits buy, at the prices on the pricing page:

• about 7 clips of 5 seconds in 720p on Wan 3.0 (33 credits per clip);
• 2 clips of 5 seconds in 720p on Seedance 2.5 (122 credits per clip);
• 31 images on Nano Banana 2 (8 credits per image);
• 125 images on GPT Image 2.5 (2 credits per image).

My advice: don't spend the bonus on Seedance 2.5 right away. Make 4–5 test clips on Wan 3.0, find a photo and a prompt that work, and only then decide whether you need a paid plan.

Paid plans

What that means in clips. By Pika's own estimate, Starter's 900 credits equal 27 clips on Wan 3.0 or 7 clips on Seedance 2.5. Creator's 3,150 credits equal 95 clips on Wan 3.0 or 25 on Seedance 2.5. For a Reel every few days, Starter is enough if your main model is Wan 3.0 or Pika 2.5.

Credit packs without a subscription

If you don't want a subscription, you can buy a pack: 60 credits per $1 on any plan, including Free. The smallest pack in the table is 1,500 credits for $25, and there's also a custom amount from $25. A subscription is the better deal: Starter gives you 900 credits for $10, while a pack would give you 600 for the same money.

A small detail that's easy to miss: top-up credits work only in the new Pika on the website and through Pika MCP. They can't be spent in the API, the iPhone app, or the old version at old.pika.art. If you want the Pikaffects from the app, a pack bought on the website won't help.

Section 03How to sign up and pay

Getting started takes a few minutes. Here's the whole path, from the first click to your first paid credits.

One detail that saves money. Before you pay, use up the bonus credits and figure out which model you like. If it turns out you need one specific model, Kling for example, it's sometimes cheaper to pay its developer directly. More on that in Kling: AI video from photos and text.

Section 04Which Pika tool for which job

Pika has more than forty apps, and a beginner drowns in them. The good news: for business you really need five or six. The rest will come in handy once you're comfortable.

Anything done with one button in an app usually costs more than the bare model: the app writes a long prompt for you and sometimes runs several models in a row. For your first tests, use Video Studio and pick the model yourself – that way you see the price of every clip.

Section 05Your first video from a photo, step by step

The most common scenario: you have a good photo and want it to come alive for 5 seconds. It's a Reels cover, a bot greeting, or a Stories background. Let's walk through an example: a photo of a nutritionist in the kitchen holding a breakfast plate.

What to expect from your first clip. In most cases the face holds up and the motion is smooth, but hands and small text on objects sometimes drift. If there's a sign or a label on packaging in the shot, be ready for the letters to warp. That's why you always end the prompt with no text and add your logo later in editing.

If you want to compare how other services handle the same task, check out how to animate a photo with AI: five services on one example.

Section 06Pikaframes: one clip from several photos

Pikaframes is one of Pika's most useful features of its own. You upload two to five frames, and the AI fills in smooth transitions between them. The result is a mini story: before and after, step one, step two, result.

Where this works for business:

• Before and after. A fitness trainer uses two photos of a client, and one flows smoothly into the other. An aesthetician shows skin before and after a course of treatments.
• Course trailer. Five slides with module topics turn into a 10-second clip for a landing page or Stories.
• Process. A nail artist shoots three stages of the work, and Pikaframes stitches them into one motion.
• Client journey. A therapist shows three scenes: an anxious person by the window, a conversation in the office, a calm walk.

How to make the transition look clean. The frames should have a similar composition: the person in roughly the same spot, the same angle, similar light. If you're full-length outdoors in one photo and in a studio close-up in the other, the AI starts inventing in-between frames, and the result looks odd. In the prompt, describe what happens between the frames: smooth transition, the person turns toward the camera, camera slowly pushes in.

Pikaframes runs on the Pika 2.5 model. If you need something longer, it's easier to make several short transitions and join them in editing.

Section 07Which model to pick so your credits last

The main beginner mistake in Pika is picking the most expensive model for every test. Then Starter's credits are gone in one evening. The price gap between models is almost fourfold, so choose for the task.

A working routine that saves credits: draft on Wan 3.0 or Pika 2.5, three or four prompt tweaks, and only the winning prompt goes to Seedance 2.5 or Veo 3.1. That way you pay for the expensive model once instead of five times.

Each of these models has its own breakdown on the blog: Seedance 2.5, Veo 3, and Kling. If you like one model so much that you make dozens of clips on it a month, compare the price directly with its developer.

Section 08The prompt formula for video from a photo

Beginners put everything into the prompt: who's in the photo, what they're wearing, the color of the walls. The AI already sees all that. What it needs is something else: what should change. A video-from-photo prompt describes motion, and only motion.

Why one or two actions. In 5 seconds a person can naturally make one movement: look up, turn, reach out. Ask them to stand up, walk to the window, pick up a cup, and wave, and the AI either speeds everything up into cartoonish fuss or skips half of it.

Camera moves that work almost every time:

• slow dolly-in – a slow push-in, adds weight and draws attention to the face;
• slow pull-back – a pull-back that reveals the setting, good for a final shot;
• camera orbits around – an orbit around the person or product, looks expensive;
• handheld camera – a slight shake, the feel of live phone footage;
• static camera – the camera stays put and only the subject moves, for talking shots.

If you want a library of ready, eye-catching camera moves at the click of a button, that's the strong suit of another service – Higgsfield. Its breakdown is in Higgsfield: how to make AI video with camera effects.

Section 0910 ready prompts for real tasks

Each prompt is designed for your own photo: upload the shot, paste the text, and change a word or two if needed. The description above each prompt tells you what it does.

1. Reels cover: a look into the camera

A photo where you're looking away or down at a notebook. A second later you raise your eyes to the viewer.

The person looks up from the notebook straight into the camera and gives a confident, warm smile. Slow dolly-in, soft natural daylight, calm mood, realistic motion. No text.

2. Nutritionist: breakfast close-up

A photo of a plate on the table. Steam rises, the camera slowly lowers toward the food.

Steam gently rises from the warm breakfast plate. The camera slowly lowers and pushes in toward the food. Soft morning window light, cozy kitchen atmosphere, realistic motion. No text, no hands.

3. Therapist: a calm office

A photo of an armchair by the window. The curtains sway, the light slowly shifts.

The curtains move slightly in a soft breeze, sunlight slowly shifts across the armchair. Very slow camera push-in, peaceful and safe mood, warm tones. No people, no text.

4. A product on the table

A photo of a jar, a book, or a box. The camera orbits the object like in an ad.

The camera slowly orbits around the product on the table. Soft studio light with gentle reflections on the surface, premium commercial look, smooth motion. No text, keep the label unchanged.

5. Fitness trainer: a power move

A photo at the gym. One short, powerful movement and a slight camera shake.

The athlete lifts the dumbbells in one powerful motion and exhales. Handheld camera with slight shake, dramatic side light in the gym, energetic mood, realistic motion. No text.

6. Tutor: the books come alive

A photo of a desk with books. The pages turn by themselves.

The pages of the open book slowly turn by themselves, dust particles float in the sunlight. Static camera, warm afternoon light, magical but realistic mood. No text, no people.

7. Nail artist: a hand in the light

A photo of a hand with a finished manicure. The hand turns smoothly, highlights glide over the nails.

The hand slowly turns toward the light, glossy reflections glide across the nails. Macro shot, static camera, soft bright studio light, elegant mood. No text, no extra fingers.

8. A greeting for your bot

A waist-up photo. You wave and nod to the viewer.

The person waves hello to the camera with a friendly smile and nods slightly. Static camera, medium shot, bright soft light, welcoming mood, natural motion. No text.

9. A course trailer with Pikaframes

Three to five slides or photos with topics. Smooth transitions between them.

Smooth cinematic transitions between the frames, the camera keeps moving forward through each scene, consistent warm light, steady pace, inspiring mood. No text distortion.

10. You and the city: the final shot

A photo outdoors or by a window with a view. The camera pulls back and reveals the open space.

The person stands still looking at the city, then turns head toward the camera. Slow pull-back revealing the wide view, golden hour light, calm confident mood, realistic motion. No text.

If a clip came out almost right, change one part of the prompt at a time: first the camera, then the light. When you change everything at once, you can't tell what actually helped.

Section 10Let Claude write the prompts

You don't have to write the prompts yourself. Tell Claude in plain words what's in the photo and what clip you want, and ask for a ready prompt based on the formula from Section 8. If you haven't used Claude yet, sign up at claude.ai. How to start from zero is covered in Learn Claude from scratch.

Here's a request you can send Claude as is:

Write a prompt for Pika, video from a photo, 5 seconds.
In the photo: me, a nutritionist, standing in the kitchen holding a breakfast plate.
What I need: I smile and hold the plate out toward the camera, the camera slowly pushes in.
This is a Reels cover.
Rules: prompt in English, 1–2 actions, camera move, light,
don't describe my appearance, end with no text. Give me 2 versions.

Notice that last tip: Claude itself suggests how to simplify the prompt if the first take fails. That saves credits better than any table.

Claude runs Pika itself through MCP

Pika has its own MCP server – a way to connect it to an assistant so the assistant can call Pika's tools itself. The server address on the official page pika.art/mcp is mcp.pika.art/mcp. It gives access to more than 60 apps and models for video, images, and sound.

What it looks like for you. In Claude's connector settings, you add a custom connector with this address, sign in to your Pika account, and grant access. Then you write in the chat: make a Reels clip from this photo, the camera slowly pushes in. Claude writes the prompt, picks the model, and runs the generation. Credits come off your Pika balance, so ask Claude right away to tell you the price before every run.

Section 11Sound, captions, and editing

A silent clip performs worse in Reels. In August 2026 Pika released four audio models, so now you can make sound in the same window:

• Pika Soundtrack – picks music, ambience, and sound effects to match the action in the shot: footsteps, clinking dishes, wind;
• Pika Music – writes a full track from a description, lyrics, or a reference;
• Pika Speech – text-to-speech with more than 50 ready voices;
• Pika SFX – a single sound effect from a description.

For captions there's Dynamic Captions: word-by-word subtitles that adjust to your speech on their own. Remember that most people scroll Reels with the sound off, so you almost always need captions.

How to put the whole clip together. A shot generated in Pika is the 5-second hook at the start of your Reel. After that comes you and your idea: phone footage, your voice, a useful tip. You can stitch it all together in any editor.

Section 12If something goes wrong

General rule: if two takes in a row come out badly, change the photo. Most often the problem is the source: dim light, cropped hands, a face that's too small.

Section 13Pika, Kling, Higgsfield, and Seedance

Pika isn't the only way to make video from a photo. Here's briefly how it differs from its neighbors, each of which has its own breakdown on the blog.

My pick for getting started: Pika, if you don't know yet which model suits you. In one evening you can try Wan 3.0, Pika 2.5, and Seedance 2.5 on the same photo and see the difference with your own eyes. Once you settle on a favorite model and your volume, compare the price with its developer.

Section 14The whole path in one minute

Start today with one action: take your waist-up photo and prompt #1 from Section 9, and make a clip with the bonus credits. Even if the first take only half works, you'll see how five seconds of motion change the impression of an ordinary photo, and you'll know whether Pika is for you.