Pika H3 Flex
1 entry about Pika H3 Flex, newest first.
Filter: categories and tools
Categories
Tools
- MiniMax H311
- ComfyUI7
- WanGP5
- FastH3-Omni4
- MOSS-Image4
- Flick2
- LTX-2.5 Layout to Render2
- Seedance2
- Adobe Premiere Generative Media Tool1
- Audio8-TTS1
Show all 80 tools
- BACH 1.0 Pro1
- Beeble SwitchX 2.01
- Boreal1
- Bulbul V4 Flash1
- Canva Craft1
- ChatGPT Images 2.51
- ComfyArt1
- ComfyUI-MangaScreen1
- ComfyUI-SceneCut1
- Concat1
- drama3-h3-talking1
- EdenLive1
- Eleven v41
- ElevenLabs Instant Soundtracks1
- Fibo 2.01
- Fish Audio Drama 31
- FLUX 3 Image Klein1
- FLUX 3 Max1
- FreeField1
- Gemini 3.8 Flash TTS1
- Gnani Timbre v3.01
- Google Vids (Gemini Omni 1.1)1
- H3 Anti-Burn LoRA1
- H3 Max1
- H3-LongTakeNoCuts1
- h3-metal1
- Henge1
- HeyGen Video1
- HeyGen Voice1
- Higgsfield Content Factory1
- Hy-Nomas1
- Ideogram Vector 1.01
- Kandinsky 6.0 Video1
- Kling 4.01
- Krea 21
- Krea 31
- LipDub 2.01
- LTX Desktop1
- Mago1
- MAI-Image-2.61
- MAI-Voice-2.11
- mflux1
- MiniMax H3 Singularity V21
- Mirelo SFX 2.01
- Nari Qwen3-TTS1
- Nvidia AI for Media1
- NVIDIA VFX Nodes1
- Ollama1
- OpenField1
- OpenFill1
- OpenKatana1
- OpenShot 4.01
- Overtone1
- P-Video-2-Pro-Avatar1
- Phonon1
- Pika1
- Pika H3 Flex1
- Plenio Music Production System1
- Qwen-Image 2.11
- ReAngle1
- Reshoot 4D1
- Runway Plugins1
- Seam1
- SolarRenderer1
- Sync Lipsync v31
- Topview 3D Builder1
- UniAVC1
- Vidu Q41
- Voxtral 31
- Whistle1
Pika H3 Flex: open-weights six-step video with sound, built on MiniMax H3, runs in ComfyUI
Pika H3 Flex is Pika's post-trained version of the open MiniMax H3 model, released with open weights on day one. It makes video with synchronized stereo audio in six sampling steps, takes up to 12 image, video and audio references, and runs locally in ComfyUI and diffusers.
Questions and answers
Is Pika H3 Flex open source?
The weights are open: Pika released two LoRA adapters for MiniMax H3 on Hugging Face, plus inference code on GitHub, and they run in ComfyUI and diffusers. They are under the MiniMax H3 Community License, which has territorial and commercial-use conditions. See the release.
How many steps does Pika H3 Flex need?
Six. Pika distilled the 50-step H3 generator to a six-step schedule; in ComfyUI use lcm with ManualSigmas and CFG 1.0. More steps do not add detail, and stacking a turbo or Lightning LoRA on top breaks the output.
How many references can Pika H3 Flex use?
Up to 12 files with the Ref2VA adapter: 9 images, 3 videos and 3 audio clips. Video and audio references must each be 2 to 15 seconds, with at most 15 seconds of each type in total.
What resolution does Pika H3 Flex make locally?
Native 480p or 768p, 5 to 15 seconds at 24 fps, with 32 kHz stereo sound. 1080p, 2K and 4K, separate audio stems and prompt enhancement are only on Pika's paid API.