The short answer
Pika is an AI creative platform from Mellis, Inc. for short video, images and sound. Since its 17 September 2026 relaunch, every plan gets every model and more than 25 apps, such as Video Studio for clips up to 30 seconds with sound. Its newest model, Pika H3 Flex, makes 15-second clips with stereo audio and is due in the app on 2 November 2026. The free plan has no monthly credits or commercial use.
| Made by | Mellis, Inc. (Pika), United States |
|---|---|
| Free plan | Yes, every model but no monthly credits and no commercial use |
| Models | Pika H3 Flex, Pika 2.5, Seedance 2.5, Wan 3.0, MiniMax H3 and more |
| Best at | Short social clips with sound, product shots, consistent characters |
| Shows sources | Not applicable: it generates video, images and audio |
| Team plans | Enterprise, with custom credits |
| Your data | Uploads may be used to improve models; no opt-out setting described |
What Pika is after the September 2026 relaunch
Pika started as a fun text-to-video app known for playful effects. On 17 September 2026 it relaunched as a single place for video, image and sound, with more than 25 focused apps sitting on top of Pika's own models and leading third-party ones. Your old login and subscription carried over, and past paid generations sit in a project called OG Pika. Pika is made by Mellis, Inc., based in Palo Alto, California.
The biggest shift is how models are offered. Pika now gives every user every model and every app, and plans differ only in credits and how many generations can run at once. The model list includes Seedance 2.5, Wan 3.0, MiniMax H3, Pika 2.5 and Pikaframes for video, GPT Image 2.5, Nano Banana 2.1 and Seedream 5 for images, and Pika's own audio models. Each app picks a suitable model for you, and you can override it.
The newest model is Pika H3 Flex, announced on 29 October 2026 and coming to the Pika creation product on 2 November 2026. It is Pika's post-trained version of MiniMax H3. It makes clips up to 15 seconds with synchronised stereo audio, accepts up to 12 reference files mixing images, video and audio, and generates natively at 480p or 768p, with Pika's upsampler taking it to 1080p, 2K or 4K. Pika has also released its generator adapters and inference code for local use under the model's licence, while the upsampler and audio stem separation stay on Pika's servers. That matters mostly to developers who want to run it themselves.
What Pika is best at
Short social clips with sound
Video Studio makes short or multi-shot videos up to 30 seconds with sound. You can attach up to 50 characters, products, settings and styles and @mention them in your prompt, which is the easiest way to keep the same mascot, presenter or product across several clips. For a weekly Reel or TikTok series, that consistency is what usually goes wrong with AI video, and Pika handles it well.
Product imagery from phone photos
Product Shot turns product photos, including ones taken on a phone, into lit campaign images, and a separate Product Ad app is aimed at short ads. For a small e-commerce brand without a studio budget, this is the most immediately useful part of Pika.
Fixing and reusing footage
Relight Media rebuilds lighting and shadows in a photo or video while keeping the subject, framing and motion. Camera Director reshoots an existing scene from new angles. Extend Video, Motion Transfer and Swap Anything cover the other common fixes. These save a reshoot when a clip is almost right.
Sound that matches the picture
Pika released four audio models on 18 August 2026. Pika Soundtrack turns a video into a synchronised mix of sound effects, voice, music and ambience. Pika Music writes tracks from text or lyrics, Pika SFX turns a description into a sound effect, and Pika Speech is a text-to-speech model that can clone a voice from a short reference clip. Silent AI clips look unfinished on social, so this closes a real gap.
Working from ChatGPT
Since 16 October 2026, a Pika plugin brings 40+ Pika apps and 100+ models into ChatGPT on desktop and web. You add the plugin, sign in to Pika, and mention @Pika in a chat. It includes a timeline for adjusting clip timing, regenerating clips and adding matching sound, and Pika adds 250 bonus credits after your first plugin generation. If you already plan campaigns in ChatGPT, this keeps the script and the clips in one conversation.
Where Pika falls short
- The free plan is a showroom. Free accounts get every model but no monthly credits. You have to buy top-up packs to create, and the free plan has no commercial licence.
- Short clips. Video Studio tops out at 30 seconds and H3 Flex at 15 seconds per generation. Longer pieces still need editing.
- A new interface. The relaunch is recent, and tutorials from before September 2026 describe an app that no longer looks the same.
- No training switch. Pika may use what you upload to improve its models, and there is no opt-out setting described.
- Not for presenter videos. For a talking avatar delivering a script, HeyGen or Synthesia are built for the job.
How to get better results from Pika
- Start in the right app. Use Product Shot for products, Relight Media for lighting fixes and Video Studio for new clips. The focused apps choose sensible models and settings for you.
- Build your cast first. Add your recurring character, product and setting as references, then @mention them in every prompt so they stay the same.
- Describe one shot clearly. Name the subject, the action, the camera move and the mood in one or two sentences. Vague prompts get generic clips.
- Add sound last. Lock the picture, then run Pika Soundtrack so the audio follows the final cut.
- Watch your parallel limit. Lower plans can run only a few generations at once. Queue your best two ideas rather than ten weak ones.
- Keep a prompt log. Note which model and prompt produced a clip you liked, so you can repeat it next week.
Privacy and your data
Pika's privacy policy, last modified on 21 October 2026, says it may use content you submit, including inputs and outputs, to improve its AI models where the law permits or with your consent. Content sent through the Pika API will generally be used to improve and train models. The policy does not describe a training opt-out setting, and privacy requests go to support@pika.art. For enterprise customers, the data and training terms in the enterprise agreement take priority. Do not upload unreleased product designs or client footage to a personal account unless you are comfortable with that.
Which Pika plan suits whom
Because every plan has every model, the choice is about volume. The free plan is for looking around and buying the odd top-up pack for personal projects. Starter adds monthly credits and commercial use, which is the minimum for any business use. Creator adds about three and a half times the credits of Starter and more parallel generations, which suits someone posting several clips a week. Fancy adds a priority queue and unlimited parallel generations for heavy users, and Enterprise covers custom volumes. Check the official pricing page for current prices.
Pika is a strong pick for social media managers, creators and small brands who want quick clips with sound and consistent characters. It is not the tool for long training videos or precise film work, where Synthesia or Luma fit better. See the best AI for video, compare tools side by side or start with our Pika lessons.
Prompts to copy
A consistent character series
Using @[character] in @[setting], create a 10-second vertical clip where [action]. Keep the character's outfit, colours and proportions exactly as in the reference. Handheld camera, natural light, upbeat mood, with matching ambient sound.
You get: A clip that matches the rest of your series
Phone photo to product ad
Turn this product photo into a 6-second vertical ad for [product]. Clean studio background in [brand colour], soft top light, slow 180-degree turn, then the product settles centre frame with space at the top for a headline. No extra objects, no text.
You get: A short product ad you can add copy to
Fix bad lighting
Relight this video as late-afternoon golden hour coming from the left window. Keep the person, their movement, the framing and the background objects exactly the same. Warm but natural skin tones, soft shadows.
You get: The same take with better light, no reshoot
New angle on an existing shot
Reshoot this scene from a low angle looking up at [subject], as if the camera were on the floor two metres away. Keep the lighting, the action and the timing identical.
You get: An alternative angle to cut between in your edit
Sound for a finished clip
Add a soundtrack to this clip: [describe scene]. I want light footsteps, a distant city hum, a soft whoosh when the camera moves, and a calm lo-fi beat underneath at low volume. No voice.
You get: A synchronised sound mix that makes the clip feel finished
Plan a week of clips in ChatGPT
@Pika I post five short videos a week for [brand] aimed at [audience]. Suggest five clip ideas with a one-line hook each, then generate the first one as a 10-second vertical clip with sound.
You get: A content plan and the first clip in one chat