Mochi 1 (preview)
A 10B-parameter text-to-video diffusion model using an Asymmetric Diffusion Transformer (AsymmDiT) architecture, released as a research preview under Apache-2.0. At release it was the largest openly licensed video generative model.
- Resolution
- 480p
- Weights
- Downloadable
- API
- No
- Commercial use
- Vendor states commercial use is permitted
Output is limited to 480p in this preview.
Apache-2.0, described by Genmo as a 'permissive Apache 2.0 license'.
Specification
- What it takes in
- text
- What it makes
- video
- Maximum duration
- not established
- Maximum resolution
- 480p
- Aspect ratios
- not established
- Controls
- text prompt, negative prompt, seed, num frames, cfg scale, LoRA fine-tuning, ComfyUI integration
- Weights
- Downloadable
- API
- No
- Pricing
- Free open weights (Apache-2.0). Self-hosting: the repository recommends about 60GB VRAM and states at least one H100 is needed for the reference implementation; bf16 with optimizations brings this to roughly 22GB, and ComfyUI community workflows report under 20GB with offloading. apiAvailable is set false: Genmo runs a hosted web playground but no first-party developer API for Mochi 1 was verified; hosted programmatic access is via Replicate and fal (third-party).
- Released or updated
- 2024-10-22
Maximum duration and supported aspect ratios are not established here. That means no page we could retrieve stated them, not that the model lacks them.
Strengths
- Fully Apache-2.0 with no restrictions
- Strong motion realism and prompt adherence for its era
- Well-supported in ComfyUI and diffusers
- LoRA fine-tuning supported
Weaknesses
- 480p only; never left preview status
- ~60GB VRAM for the reference implementation
- Clearly outclassed by Wan 2.2, HunyuanVideo 1.5 and LTX-2
- Inconsistent documentation of maximum clip length
- Genmo has not shipped a Mochi 2
Best for
- Research, architecture study and permissive-license fine-tuning experiments rather than production output.
Where the sources disagree
Recorded rather than resolved. Picking the tidier number would hide the fact that the vendor's own pages do not agree.
- Max frames/duration conflict: 31 frames (repo example), 84 frames (Hugging Face card), and ~5.4s at 30 fps / 163 frames (Genmo launch materials). Recorded as null.
What to understand before using it
Explainers naming the mechanisms this model's own documentation mentions.
See it from the vendor
Shotcraft hosts no sample output. Judging a model means watching its own showreel with your own eyes, and a clip re-hosted here would be a clip we chose.
Nearby in the directory
Adobe Firefly Video Model
The Adobe Firefly Video Model generates video from a text prompt, optionally guided by an image used as the first or final frame. Adobe's Firefly Serv
Video generation
Allegro
A small text-to-video model comprising a 2.8B VideoDiT and a 175M VideoVAE, released under Apache-2.0 in October 2024. It generates 88-frame clips at
Video generation
Amazon Nova Reel 1.1
Amazon Nova Reel generates video from text prompts or from text plus an image, and is available only through Amazon Bedrock. AWS documents output at 1
Video generation