ZMIME
Video model comparison · 2 models · Updated Nov 21, 2026

H3-Tiny-17B-FewStep-Ref2VA-v0.1 vs AURA 1.3B

Newest: AURA 1.3B · Native audio: H3-Tiny-17B-FewStep-Ref2VA-v0.1. Not enough arena votes yet for a quality score, so this is a spec comparison.

  1. H3-Tiny

    H3-Tiny-17B-FewStep-Ref2VA-v0.1

    Released Nov 19, 2026

    • Text to video
    • Image to video
    • Native audio
    • Arena—
    • From—
    • Max length—
  2. Kyutai

    AURA 1.3B

    Released Nov 20, 2026

    • Image to video
    • Audio input
    • Arena—
    • From—
    • Max length—
  3. Add a model

    Make it a three-way comparison.

01 — At a glance

Where they differ

Quality is the arena score from blind, side-by-side human votes. Price, length and resolution come from each lab’s own listings.

  • NewestAURA 1.3BH3-Tiny-17B-FewStep-Ref2VA-v0.1 Nov 19, 2026 · AURA 1.3B Nov 20, 2026
  • Widest inputsSame inputsH3-Tiny-17B-FewStep-Ref2VA-v0.1: Text prompts, Reference images · AURA 1.3B: Reference images, Audio
  • Native audioH3-Tiny-17B-FewStep-Ref2VA-v0.1Generates an audio track with the video: H3-Tiny-17B-FewStep-Ref2VA-v0.1 yes · AURA 1.3B unconfirmed
  • Image to videoBoth animate imagesStarts from an uploaded image: H3-Tiny-17B-FewStep-Ref2VA-v0.1 yes · AURA 1.3B yes
  • Self-hostingAll open weightsDownloadable weights (MiniMax H3 Community License and CC BY-NC 4.0)
  • Most API hostsSame reachH3-Tiny-17B-FewStep-Ref2VA-v0.1: 0 · AURA 1.3B: 0
02 — Side by side

Every spec in one table

Dashes mean the data is not published.

H3-Tiny-17B-FewStep-Ref2VA-v0.1 vs AURA 1.3B specifications side by side
SpecificationH3-Tiny-17B-FewStep-Ref2VA-v0.1H3-TinyAURA 1.3BKyutai
TaskAudiovisual generationSpeech-driven portrait animation
Published capabilitiesGenerates video and synchronized sound from reference imagery using a pruned MiniMax H3 transformer.A four-step causal model animates a portrait from streaming speech, rendering chunks from the available audio.
Model size17.08B transformer parameters1.3B parameters
AccessDownloadable transformer checkpoint for the H3 pipelinePublic student checkpoints and inference code
Operating limitsSix steps by defaultKyutai reports 20.7 fps end to end on one H100 at 576p for 16 Hz playback.
SourcesModel weightsModel weights
Output
ProducesVideo with soundVideo
Resolutions—576p
Highest resolution—
SoundAlways generated—
Inputs
Text promptsYesNo
Start image (image to video)YesYes
Video inputNoNo
Audio inputNoYes
Access
WeightsOpenMiniMax H3 Community LicenseOpenCC BY-NC 4.0
ReleasedNov 19, 2026Nov 20, 2026
03 — Questions

Which should you choose?

Which is better, H3-Tiny-17B-FewStep-Ref2VA-v0.1 or AURA 1.3B?

Neither model has enough arena votes for a reliable quality score yet, so this page compares specs only. Run the same three or four prompts through each and judge the results.

What is the difference between H3-Tiny-17B-FewStep-Ref2VA-v0.1 and AURA 1.3B?

H3-Tiny-17B-FewStep-Ref2VA-v0.1 is a video model from H3-Tiny (Nov 19, 2026), taking text prompts and reference images and producing video with sound. AURA 1.3B is a video model from Kyutai (Nov 20, 2026), taking reference images and audio and producing video.

Which is newer?

H3-Tiny-17B-FewStep-Ref2VA-v0.1 Nov 19, 2026; AURA 1.3B Nov 20, 2026.

Which generates video with sound?

Generates an audio track with the video: H3-Tiny-17B-FewStep-Ref2VA-v0.1 yes; AURA 1.3B unconfirmed.

Can I run any of these on my own hardware?

All open weights. Downloadable weights (MiniMax H3 Community License and CC BY-NC 4.0).