H3 Flex AI Video Generator: Guide, Specs & Early Access
A practical guide to Pika’s new H3 Flex video model — what it does, specs, prompt tips — plus a simple H3 Flex generator coming soon. Join the waitlist.
Status: H3 Flex generation on this site is not live yet. We are building a simple text-to-video and image-to-video generator for the H3 Flex model. Nothing below charges you or generates video today.
Leave your email and we will send one message when it opens (launch pricing from $9.9 per video, no subscription).
Want to try H3 Flex right now? Pika runs it in its own app and developer API; the open weights are on Hugging Face.
Video + audio together
H3 Flex outputs MP4 with 32 kHz stereo audio mixed in — dialogue, music, effects.
5–15 seconds
Integer durations from 5 to 15 s at 24 fps; 11 aspect ratios including 9:16.
Reference-guided
Use first/last frames or up to 12 reference files to keep characters and products consistent.
H3 Flex at a glance
- What it is: Pika H3 Flex, a post-trained version of MiniMax H3 that generates video with synchronized stereo audio in six sampling steps (model card).
- Length: 5–15 seconds per clip, 24 fps; resolution: native 480p / 768p, with 1080p, 2K and 4K via Pika's API upscaler (API docs).
- Modes: text to video, first-frame / last-frame / first-and-last-frame, and multimodal references (up to 9 images, 3 videos, 3 audio clips).
- Aspect ratios: 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16, 9:21.
How to use H3 Flex
- Pick a mode. Text to video for a new scene; first-frame (image to video) when a character, product or layout must match a picture; references when you need several images, a motion video or audio to guide it.
- Write a timed prompt. Describe subject and setting, then the camera move for each time range (e.g. 0–3 s, 3–6 s), the playback speed, and a separate Sound line.
- Choose length, ratio and resolution. Request 5–15 seconds, an aspect ratio such as 16:9 or 9:16, and 480p/768p (native) or an upscaled 1080p–4K on Pika’s API.
- Generate and iterate. Generate, review, then change one thing at a time — camera, timing or sound — and keep the seed to compare.
H3 Flex launch pricing (planned)
Payments are not open yet — these are the prices we plan to launch with. Join the waitlist to be notified.
No subscription. Prices may change before launch.
H3 Flex FAQ
What is H3 Flex?
H3 Flex (Pika H3 Flex) is an open-weight AI video model from Pika, published on Hugging Face in October 2026. It is post-trained from MiniMax H3 and generates 5–15 second videos with synchronized stereo audio from text, start/end frames or reference files.
How do I write a good H3 Flex prompt?
Describe the subject, the setting, the camera move for each time range (for example 0–3 s, 3–6 s), the playback speed, and the sound you want. See our H3 Flex guide for a template.
Can I generate H3 Flex videos on this site today?
Not yet. Generation on this site is not live. Join the waitlist and we will email you once it opens. Planned launch pricing is $9.9 for one video, $19.9 for three or $29.9 for five, with no subscription.
Where can I use H3 Flex right now?
Pika offers H3 Flex through its own app and its developer API (dev.pika.art), and the open weights are on Hugging Face (Pika-Labs/Pika-H3-Flex) for local GPUs.
Text to video or image to video — which should I use?
Use text to video for new scenes. Use image to video when you need a specific character, product or composition in the first frame.
Is this the official Pika website?
No. H3Flex Studio is an independent site, not affiliated with Pika or MiniMax. The official H3 Flex pages are at pika.art, dev.pika.art and huggingface.co/Pika-Labs.
What license are the H3 Flex weights under?
Pika’s model card says the H3 Flex adapters are model derivatives of MiniMax H3 and are distributed under the MiniMax H3 Community License Agreement. Read that license before any commercial or local use.