Pika-H3-Flex

Pika H3 Flex

Built for professional iteration.

Six-step generation · Omni-reference · Native 480p / 768p · Up to 4K via Pika API

Pika API Create with Pika Tech Blog GitHub Hugging Face

Model weights · Watch samples · Quickstart · Prompt guide · Pika API extensions

Pika H3 Flex is a post-trained H3 model, designed to give professional creators the flexibility and speed to explore with precision. Optimized for greater control, it generates video with synchronized audio in six sampling steps and supports up to 12 reference inputs, with shot direction through references and prompts.

Pika’s hosted API extends this workflow with 1080p, 2K and 4K video generation, audio stems and prompt enhancement. Local open-weight inference supports native 480p / 768p video with mixed stereo audio.

Ideal for multi-shot storytelling · Motion graphics · Product films · Reference-guided animation

Watch the model

Explore camera choreography, cinematic storytelling, graphic design and reference-guided animation. Play a clip with sound; expand its prompt for the generation description.

Prompts and input assets · Showcase metadata

These are creative showcases, not release-validation renders or throughput benchmarks. Exact reproduction with the released package is not guaranteed.

System overview

Pika H3 Flex is one video-and-audio model family with two task adapters: FL2VA for text and endpoint-frame conditioning, and Ref2VA for multimodal references. It builds on MiniMax-H3 and generates video with synchronized stereo audio in six sampling steps.

The inference pipeline encodes the prompt and input media, generates video and audio latents with the selected Pika H3 Flex adapter, then decodes and combines them into an MP4. The package downloads the pinned base components and adapter automatically. Use the prompt guide to describe scenes, motion, dialogue, ambience and music; optional prompt enhancement is available through the Pika API.

Input and output specifications

Durations are rounded up to the supported frame grid: a 5-second request produces 124 frames (about 5.17 seconds), and a 15-second request produces 362 frames (about 15.08 seconds). Native resolution and aspect ratio jointly determine the actual canvas dimensions.

Generation modes and reference limits

Ref2VA accepts up to 9 images, 3 videos and 3 standalone audio clips, with 12 reference files total. At least one image or video is required. Each video or standalone audio clip must be 2–15 seconds; combined video duration and combined standalone-audio duration must each be at most 15 seconds. A video's embedded soundtrack stays with that video and does not count as an additional standalone audio file.

These are accepted input limits, not a guarantee that every maximum-size combination fits GPU memory. The package runs one request on one GPU, with optional CPU offload. See hardware requirements and tested configurations for representative H100 results and memory guidance.

Pika API extensions

Extend your generation workflow with Pika’s hosted endpoints:

These optional hosted services are separate from the downloadable adapters and local inference package. The specifications above describe local model inference; see each endpoint’s documentation for its API capabilities and usage.

Run it

Clone the inference repository and follow its pip, Conda, or Docker quickstart. The pika-h3-pro command downloads the pinned base components and selected adapter. See the prompt guide for structured prompts.

Weight packages

The repository also includes Diffusers-format MiniMax-H3 base components: transformer/, transformer_ref/, text_encoder/, vae/, audio_vae/, tokenizer/processor assets and scheduler configurations, assembled by modular_model_index.json. These transformer checkpoints are base models; apply the matching Pika adapter separately. The Diffusers examples perform both loading steps explicitly. Base components occupy approximately 210 GB in total; a selected workflow loads only its matching transformer partition.

Each manifest records the public release ID, pinned base revision, adapter checksum, conditioning precision and exact six-step sampling schedule. Generic six-step Euler sampling is not equivalent. Base components: MiniMaxAI/MiniMax-H3.

Other inference platforms

• ComfyUI — See the Comfy tutorial; use the R2V workflow, FL2VA workflow or T2V workflow. Comfy-format BF16 LoRA weights are included in comfy/loras/; see the adapter guide for downloads and checksums.
• Hugging Face Diffusers — See the official Pika H3 Flex documentation and our installation and generation examples for FL2VA first-frame conditioning and Ref2VA image/video references. The examples use the public upstream Diffusers commit pinned in examples/requirements-diffusers.txt; this installation is separate from the Pika CLI above.

Hardware and examples

The Pika CLI inference package runs one request on one GPU. Representative 480p and 768p CLI requests passed on H100 80GB. Large requests may require --cpu-offload and substantial host RAM. See hardware requirements and tested configurations. The separate Diffusers examples have their own installation and validation notes in examples/README.md.

The featured showcase manifest includes prompts, input assets and video hashes. Seeds, exact checkpoint revisions and service preprocessing for these examples are unspecified.

License

These adapters are Model Derivatives of MiniMax H3 and are distributed under the MiniMax H3 Community License Agreement.

See NOTICE for the required MiniMax notice and Mellis, Inc. copyright statement.

Bundled MiniMax-H3 base components retain the MiniMax agreement. The Qwen encoder and associated tokenizer/processor assets are covered by Apache 2.0; see the component sources and notices. The example Python code is covered by Apache 2.0; that code license does not cover model weights or example media.

Attribution

If you use these adapters in research or a product, we'd appreciate a credit to Pika and a citation:

@misc{pika2026flex,
  title  = {Pika H3 Flex Technical Deep Dive},
  author = {{Pika Research}},
  year   = {2026},
  howpublished = {\url{https://pika.art/blog/pika-h3-flex-research}}
}