GoogleFreeDeveloperZhipu AI
ReleasedCogVideoX 1.5 · Nov 2024
Native audioAdd voiceover
Max resolution1360×768
Max clip length6–10s
Pricing fromFree (open) · API
Key specs
| Developer | Zhipu AI |
| Origin | China |
| Released | CogVideoX 1.5 · Nov 2024 |
| Native audio | Add voiceover |
| Max resolution | 1360×768 |
| Max clip length | 6–10s |
| Availability | Open weights · Open weights (2B: Apache 2.0) |
| Available via | Open weights · Z.ai API |
| Pricing from | Free (open) · API |
Where CogVideoX shines
- Runs on modest consumer GPUs (as little as ~4–5 GB VRAM)
- Broad ecosystem — Diffusers, ComfyUI, LoRA
- The 2B model is Apache-2.0 and commercially free
- Influential, research-grade architecture
Best for
CogVideoX is best for quick, flexible text- and image-to-video where an accessible, open model is a good fit.
Local short-clip text- and image-to-videoResearch and fine-tuningCommercial use on the permissive 2B tierPrototyping and B-roll
Compare CogVideoX with other models Good to know
- !Short clips and lower resolution than newer models
- !No native audio in the open weights
- !English-only prompts; open releases are now dated
Background. From Zhipu AI and Tsinghua's THUDM, CogVideoX (1.5, November 2024) is one of the most-forked early open video models, runnable on consumer hardware with an ICLR-accepted design.
How to use CogVideoX on Vivideo
1
Open Vivideo and start a new video.
2
Select CogVideoX as your model — or let the Video Agent pick it for you.
3
Describe your shot (or upload an image) and set duration and aspect ratio.
4
Generate with CogVideoX, then refine, add a voice or avatar, and export for any platform.
Every top model, one subscription
Each model is one of 30+ in Vivideo — switch per shot to get exactly the look you want.
GoogleFree
Google
OpenAI
OpenAI
ByteDanceSeedance 2.0
State-of-the-art motion, realism and native audio.
AudioUltra-wide
Generate with Seedance 2.0 
Kling
Kling
AlibabaAbout CogVideoX on Vivideo
CogVideoX, from Zhipu AI, is a widely used open-source video model supporting text- and image-to-video. Its openness has made it a community favorite for experimentation.
CogVideoX suits creators who want a flexible, accessible engine for everyday clips and quick iterations across both text and image inputs.
On Vivideo, CogVideoX is one of 30+ models on a single subscription — use it as a flexible everyday option, then escalate to Veo or Sora when a shot needs extra polish, all in one project.
CogVideoX earned its place as one of the most-forked open video models: the 2B variant is Apache 2.0 and commercially free, it runs on as little as ~4–5 GB of VRAM, and the Diffusers, ComfyUI and LoRA ecosystem around it runs deep. CogVideoX text to video and image to video both work well for quick, everyday clips, which made it the default starting point for a generation of tinkerers and researchers.
Be realistic about its age: the open releases date to late 2024, resolution tops out at 1360×768, clips run 6–10 seconds, there's no native audio, and prompts are English-only. Against newer engines it trails on fidelity — WAN and Hunyuan are the stronger open picks today, and Veo 3.1 or Sora 2 win outright on polish. It remains a fine choice for prototyping, B-roll and research.
On Vivideo, a CogVideoX AI video is free to try — daily free generations, no local GPU setup. Use it for fast drafts and layout passes, then re-render the keepers on a heavier model; since output is silent, add an AI voiceover and music in the same project before exporting, or fold clips into a longer 10-minute edit.
CogVideoX FAQ
Is CogVideoX free on Vivideo?
You can try CogVideoX free on Vivideo to start — no credit card. Heavier use and premium models are covered by a paid plan.
What is CogVideoX best at?
CogVideoX is an accessible open model that handles both text and image inputs well, making it a flexible everyday option.
Does CogVideoX support text and image to video?
Yes — on Vivideo you can use CogVideoX for both text-to-video and image-to-video, and switch to other models per shot.
Can I switch from CogVideoX to another model mid-project?
Yes — Vivideo lets you switch models per shot, so you can mix CogVideoX with other engines in one project.
Is CogVideoX free?
The 2B open weights are Apache 2.0 — free including commercial use if you self-host. On Vivideo you can also try it free, with daily free generations and no setup.
What resolution does CogVideoX support?
Open releases top out at 1360×768. For 1080p or 4K delivery, generate the shot with a newer model like Veo 3.1 or LTX-2 instead.