Home / Video · Image / Pika H3 Flex

Pika H3 Flex
For image-to-video creation: a 33B open-weight model you can use in the cloud or self-host
What is Pika H3 Flex
Pika H3 Flex is a 33B-parameter open-weight image-to-video model that can be used in Pika's cloud or self-hosted. Launch pricing starts at $0.01 per second with 15 free generations per day, and the weights are released under the minimax-h3-community license on Hugging Face.
How to use
- Try Pika H3 Flex in Pika (create.pika.art) or via the Pika API playground at dev.pika.art.
- Choose a mode: text to video, first/last-frame image to video, multimodal reference to video, or prompt + reference-guided video editing.
- Add a prompt and up to 12 references (sketch, voice sample, image, sound, video, moodboard).
- Draft quickly at 480p or 768p, then finish the same take at 1080p, 2K or 4K with draft-to-video.
- Optionally export separate Dialogue, Music and Effect audio stems as WAV files.
- To run locally, download the open weights from Hugging Face and the code from GitHub (Python 3.12 and an NVIDIA GPU;
- native 480p/768p).
Use cases
- Fast iteration on video ideas at draft resolution before finishing in 4K Complex camera moves directed in natural language Multi-shot 15-second sequences with dialogue, music and sound design Motion graphics with controlled typography and transitions Product films that keep products recognizable Animated shorts with consistent characters and native audio
FAQ
What is Pika?
Pika is the creative platform that lets you create and edit videos, images, and sound using AI, with creative apps for everything from product photos to talking characters. Make something from scratch or transform what you already have.
How does Pika work?
Choose an app for what you want to make, add a prompt or your own media, and generate. Pika's apps use Pika's own models and models from other providers. Some apps choose the model for you; others let you pick your favorite and adjust the settings.
Do you have an API?
Yes. The Pika API Club gives developers access to 100+ of the leading video, image, and audio models through one API. API billing is separate from the Pika product.
Can I choose the model myself?
Yes. In apps with a model selector, you can choose the model you want to use. The available settings depend on that model, including the references you can add, video length, resolution, and sound. Some apps take care of the model choice for you.
Price from official sources
Freemium
- Pika H3 Flex API prices are charged per output-video second (regular / launch price): Text, image and keyframe generation: 480p $0.02 / $0.01;
- 768p $0.04 / $0.02;
- 1080p $0.07 / $0.035;
- 2K $0.08 / $0.04;
- 4K $0.16 / $0.08.
- Reference to video (input + output seconds when a video reference is used): 480p $0.06 / $0.03;
- 768p $0.12 / $0.06;
- 1080p $0.15 / $0.075;
- 2K $0.16 / $0.08;
- 4K $0.20 / $0.10.
- Image and audio references are free.
- Audio stems:$0.001/second.
- Prompt rewriter API (for self-hosting): $0.02/request (free on Pika-hosted).
- Open weights are free to download.
- Pika app plans:Free;
- Starter $10/mo;
- Creator $35/mo;
- Fancy $95/mo;
- Enterprise custom.
Check the official site for current pricing · Price not yet checked by our editors
Source: pika.art · Info as of Nov 2, 2026
Company
- Name
- Mellis, Inc.
- Address
- 849 High St, Palo Alto, California 94301, USA
Reviews
No reviews yet.
Alternatives
-
LFM3-Omni-2B (Liquid AI)
Free
For on-device multimodal apps: a 2B omni-modal model supporting speech, images, and video
- Added
- Nov 2, 2026
Same category: Video
-
Pause 1.0 Pro
Paid
For AI video editing: describe changes in text to edit an entire video in one pass
- Added
- Nov 2, 2026
Same category: Video
-
Riverside Agents
Freemium
For podcast and video production: AI agents handle prep, recording, editing, and publishing
- Added
- Nov 2, 2026
Same category: Video
-
Unlimited-OCR-2 (Baidu)
Free
For document recognition: Baidu's open-source 3B-parameter OCR model that can be deployed locally
- Added
- Nov 2, 2026
Same category: Video
Social