Patreon
Home / Videos / Pika H3 Flex Is Here: Cinematic AI Video With Its Own Sound,

Pika H3 Flex Is Here: Cinematic AI Video With Its Own Sound, Running Locally on Your PC

Pika H3 Flex from Pika Labs now runs in ComfyUI and SwarmUI presets: a 5-second 1344 x 768 clip with stereo sound takes about a minute on an RTX 5090.

Published Oct 30, 2026 · 2,225 views · Watch on YouTube

Get the app and files

Pika H3 Flex Presets for ComfyUI & SwarmUI: Video With Synced Sound

Pika H3 Flex Presets for ComfyUI & SwarmUI: Video With Synced Sound

Pika H3 Flex from Pika Labs now runs in ComfyUI and SwarmUI presets: a 5-second 1344 x 768 clip with stereo sound takes about a minute on an RTX 5090.

Chapters

  1. 0:00 A ship sails out of its bottle in 6 steps
  2. 0:10 One text prompt makes a full scene with sound
  3. 0:20 Two reference images become a talking captain scene
  4. 0:31 All ComfyUI V147 presets and the three new ones
  5. 0:44 All 94 SwarmUI presets with the new Pika cards
  6. 0:57 Download every model and run the presets in both apps
  7. 1:15 Pika H3 Flex results from real RTX 5090 runs

About this video

Pika H3 Flex, the new open model from Pika Labs built on MiniMax H3, makes cinematic video with synchronized sound in only 6 steps, and it now runs in our ComfyUI and SwarmUI presets. In this 90-second showcase every clip is a real run on an RTX 5090: two images become one continuous shot as a ship sails out of its bottle onto a stormy ocean, one text prompt gives a blacksmith, his golden eagle, sparks, steam, dialogue and music, and a character image plus a place image become a talking airship captain scene. A 5-second 1344 x 768 clip with stereo sound takes about a minute.

Links:

ComfyUI installer and presets, including the 3 new Pika H3 Flex presets (V147): [ https://www.patreon.com/posts/105023709 ]

SwarmUI installer, presets and model downloader with the MiniMax-H3 Core Bundle (V187): [ https://www.patreon.com/posts/swarmui-auto-and-114517862 ]

One-time Windows requirements installer tutorial: [ https://youtu.be/ptuLL0C72Ms ]

SECourses Discord: [ https://discord.com/invite/software-engineering-courses-secourses-772774097734074388 ]

You will see the three new presets in both apps: Text To Video, First + Last Frame To Video and References To Video. Each switches by itself between the MiniMax H3 text and frames model and the references model, and keeps prompt batches, folder batches, Init Audio, an optional LoRA stack and face refinement. The MiniMax-H3 Core Bundle of our model downloader gets every file in one click; if you already have it, only the two Pika H3 Flex LoRAs download. In ComfyUI, open Workflows, SECourses and pick a Pika H3 Flex preset: Text To Video needs only your prompt; First Last Frame To Video takes your first image in INIT IMAGE and your last image in LAST FRAME and adds the timing line Pika H3 Flex expects with the real clip length; References To Video takes up to 9 images, 3 videos and 3 audio files with the + Add references button and @image1, @video1 and @audio1 tokens in the prompt. In SwarmUI, apply the preset from the Presets tab: Text to Video, Beginning and Ending Frame to Video (Init Image plus Video End Image under Display Advanced Options) or References to Video with the images attached beside the prompt.

Main topics include Pika H3 Flex, MiniMax H3, 6-step distilled video generation, text to video with audio, first and last frame to video, image references to video, ComfyUI presets, SwarmUI presets and the one-click MiniMax-H3 Core Bundle. The presets keep Pika's official recipe: LoRA strength 1.0, the lcm sampler, CFG 1 and Pika's 6-step schedule, and they output 24 FPS video with 32 kHz stereo audio at 768p in every aspect ratio, 5 to 15 seconds long. Our MiniMax H3 prompt guide for ChatGPT, Claude and Gemini now covers Pika H3 Flex, so an LLM can write the structured prompts with the scene, the camera moves, the spoken lines, the sound effects and the music for you.

Use the chapters to jump to each Pika H3 Flex mode, the preset collections, the model download and the results.

Chapters:

0:00 A ship sails out of its bottle in 6 steps

0:10 One text prompt makes a full scene with sound

0:20 Two reference images become a talking captain scene

0:31 All ComfyUI V147 presets and the three new ones

0:44 All 94 SwarmUI presets with the new Pika cards

0:57 Download every model and run the presets in both apps

1:15 Pika H3 Flex results from real RTX 5090 runs

Music: "PATCH SKETCH 019 - Another Liquid Drum and Bass Loop" by thehorsevalse (Freesound).

This showcase is for anyone who wants cinematic AI video with real sound on their own PC. Download the latest ComfyUI V147 or SwarmUI V187 package from the links above, get the MiniMax-H3 Core Bundle with one click and open the Pika H3 Flex presets.

How to get it: download ComfyUI_V147.zip and run Windows_Install_Or_Update_ComfyUI.bat, or download SwarmUI_Model_Downloader_v187.zip, run Windows_Install_or_Update_SwarmUI.bat and import the presets with Windows_Preset_Delete_Import.bat. Then start the model downloader with Windows_Start_Download_Models_App.bat and click Download All Models on the MiniMax-H3 Core Bundle; files you already have are checked and kept.

For support, questions, feature requests, installation issues or future updates, check the links in this description, the pinned comment, Patreon, Discord and the video comments. Thank you for watching.