Veo 3.1
Generate short videos with native audio from text prompts, images and visual references.
Pika H3 Flex
pika.art
Generate short videos with synchronized audio using text, endpoint images, and mixed media references.
Visit website
Pricing
Paid
Available on
Web playground, API, Self-hosted inference
Category
AI Video Tools
What is Pika H3 Flex?
Pika H3 Flex is a video generation model post-trained on MiniMax H3 for creators and teams developing short visual sequences. It accepts text prompts, first or last frames, and combinations of image, video, and audio references. Reference-guided editing lets creators direct changes to existing footage.
Hosted access is available through Pika API Club, while downloadable generator adapters and inference code support local workflows. The hosted service adds high-resolution finishing and separate audio stems. Availability in the Pika creation product is announced for November 2, 2026.
Key features
Multiple generation modes
Create from text, condition a clip on first or last images, combine media references, or use prompts and references to guide video edits.
Synchronized video and sound
Generate native 480p or 768p video with synchronized stereo audio. Requests support 5–15 seconds, with output duration rounded to the model's frame grid.
Hosted finishing and audio stems
Upscale clips to 1080p, 2K, or 4K. Optional Dialogue, Score, SFX, Foley, and Ambience stems let editors adjust soundtrack components separately.
Downloadable generation components
Generator adapters and inference code support text, endpoint-frame, and multimodal reference generation with audio. Local performance depends on hardware and setup.
Who is it for?
Advertising concept exploration
Creative teams can compare short product-video directions before selecting a concept for further editing.
Social clip production
Video creators can develop short sequences from prompts and reference imagery, then refine the chosen soundtrack.
Generation workflow integration
Developers can connect asynchronous video generation to an existing creative application.
How to use Pika H3 Flex
- Sign in to Pika API Club and create a project API key.
- Choose the H3 Flex endpoint for your input type and review its parameters and pricing.
- Prepare a prompt and any references; make reference files accessible through hosted URLs.
- Submit a generation request, poll its job status, and retrieve the completed output URL.
Pika H3 Flex pricing
Paid
Hosted generation is billed per output second, with charges only for successful runs. Rates vary by endpoint and resolution; reference-video workflows also charge for input-video seconds. A temporary promotion gives Pika API Club members 15 free H3 Flex generations daily from October 29 through November 5, 2026. See the official pricing page for current rates and launch offers.
Check current pricing ↗What to know before you use it
- Reference mode requires an image or video; audio-only references are unsupported. It accepts at most 12 assets: up to nine images, three videos, and three audio clips. Each video or audio reference must last 2–15 seconds, with combined limits of 15 seconds for video and 15 seconds for audio.
- Local inference requires Python 3.12, an NVIDIA GPU, and a compatible driver. Representative tests use one H100 80 GB per request; large requests can exceed memory, and a single request cannot span multiple GPUs.
- Upscaling, stem separation, prompt enhancement, and accelerated serving are hosted features. Downloadable adapters use the MiniMax H3 Community License Agreement.
- The API documentation excludes frame-exact rotoscoping, feature-length rendering, pixel-perfect compositing, and precise on-screen text from its intended tasks.
Frequently asked questions
Does H3 Flex generate native 4K video?
No. Native generation is 480p or 768p; hosted finishing upscales the result to higher resolutions, including 4K.
Can I use it without a provider-specific SDK?
Yes. The API accepts HTTPS requests with JSON and returns asynchronous jobs that can be polled for output URLs.
Can I edit the generated soundtrack separately?
The hosted API can return five separate audio stems for mixing dialogue, music, effects, Foley, and ambience.