Helios-Base — Video Generation AI Model

What it does

Helios Base is the highest-quality variant of the Helios video family from PKU-YuanGroup. It generates long-form video from a text description, a reference image, or an existing clip.

The distilled variant trades quality for speed; Base is the one to reach for when the output matters more than the wait, and it is the most detailed and coherent of the family.

When to use it

• Video work where output quality matters more than turnaround
• Generating from a text description, or from a single seed image
• Longer clips that have to stay coherent rather than drifting
• Running on a single GPU instead of a multi-GPU cluster
• Dropping into an existing production pipeline without rework

Details

• CategoryVideo Generation
• Year2026
• LicenseApache-2.0

Compliance & Provenance

• ProviderOpen-source (BestWishYsh) · Specialized
• EU AI Act RiskLimited Risk
• Art. 50 TransparencyRequired — AI-generated outputs are marked

Inputs & Outputs

TextInput · string

Text prompt for video generation

ImageInput · image · optional

Optional input image for image-to-video generation

VideoOutput · video

Generated video

Parameters

resolutionstring · default 640x384 (5:3) · options: 768x320 (12:5), 768x384 (2:1), 640x384 (5:3), 768x512 (3:2), +7 more

Resolution (W×H, ratio)

Output resolution with aspect-ratio label. Landscape options: '768x320 (12:5)', '768x384 (2:1)', '640x384 (5:3)' (default), '768x512 (3:2)', '576x448 (9:7)'. Square: '512x512 (1:1)'. Portrait: '448x576 (7:9)', '512x768 (2:3)', '384x640 (3:5)', '384x768 (1:2)', '320x768 (5:12)'. Higher resolution increases VRAM and latency.

num_framesinteger · default 99 frames (~4s @ 24fps) · options: 99 frames (~4s @ 24fps), 198 frames (~8s @ 24fps), +13 more

Frame count (chunk multiples)

Total frames to generate. Must be a multiple of 33 per chunk. Available presets from '99 frames (~4s @ 24fps)' up to '1452 frames (~60.5s @ 24fps)'. Longer clips are proportionally more expensive in time and VRAM. Default ~4 seconds.

fpsinteger · default 24 · options: 16, 24

Output FPS (16/24)

Output frame rate. 16 = lower file size, smoother for stylized/animated content. 24 (default) = cinematic standard.

stepsinteger · default 50 · range 20–60 · advanced

Diffusion steps (20-60)

Number of denoising steps. Default 50 — full quality. Lower values (20-30) save time at moderate quality cost; very low (<20) noticeably degrade fidelity. Bump up to 60 only if quality is unsatisfactory at default.

seedinteger · default -1 · range -1–2147483647 · advanced

Random seed (-1 = random)

Random seed for reproducibility. -1 (default) selects a fresh random seed every run. Set a fixed non-negative integer to reproduce a previous video (same seed + same prompt + same params → same output). Useful for iterative prompt refinement.

Tags

• video-generation
• text-to-video
• image-to-video
• diffusion

Alternatives in Video Generation

• Cosmos3 Nano

  NVIDIA world model for text-, image- and video-to-video, with optional synced audio.

• Helios Distilled

  Fast distilled text-to-video, fixed 640x384. 1-5 min.

• LatentSync 1.6 (Lip Sync)

  Re-sync a talking video's lips to a new voice.

• LTX-2.5

  Lightricks LTX-2.5 distilled 22B: text, or text plus a first-frame image, to a video with a synchronized soundtrack generated alongside it.

• LTX-2.5 Audio-to-Video

  Make a talking video from one picture and a voice.

• MiniMax-H3 Ref2VA

  MiniMax-H3 Ref2VA (omni-reference) generates a 5-15s 24fps video with native stereo audio from a prompt plus a reference image and a reference clip, that clip's own soundtrack included. Video and its soundtrack come out of one denoising loop, so lip movement and sound land in sync. ~144GB bf16 weights — offloaded component-by-component, minutes-scale per clip.

• Wan2.2 TI2V 5B

  Wan2.2 dense 5B unified text+image-to-video. Text only → 5s video from prompt; text + image → image as the first frame of the video.

Resources

• github
• huggingface
• paper