MiniMax H3
Ranked 2 of 9 AI video generators by recommenders
Open-weight AI video model that people run locally to make short clips quickly. Critics find results inconsistent, often bland, and behind rival video models.
Open-weight AI video model that people run locally to make short clips quickly. Critics find results inconsistent, often bland, and behind rival video models.
40 people gave an opinion
- recommend it
- 29
- criticize it
- 11
- approval share
- 73%
Recommended in 3 of 13 months
- November 2025: not recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: not recommended
- April 2026: not recommended
- May 2026: not recommended
- June 2026: not recommended
- July 2026: not recommended
- August 2026: recommended
- September 2026: recommended
- October 2026: recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 29 people recommend itNeeds 10 or more, met.
- 2.6 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 3 of 13 monthsNeeds 4 or more, not met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Runs locally on consumer GPUs
“If you want a good balance of size, speed and quality you should use the int8_convrot model”
Comment from 11 Aug 2026 in “Antirez/h3.c: MiniMax H3 inference engine for Mac computers” “The int8 release is very capable and runs acceptably on consumer gpus”
Comment from 28 Aug 2026 in “Gemini Omni 1.1 Flash” “Minimax H3 running local can generate amazing short vids quite fast”
Comment from 10 Aug 2026 in “DeepSeek costs OpenCode Go user $1.14/day; dual DGX breaks even in 24 years” - Comment from 4 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video”
- Comment from 10 Aug 2026 in “Meta Muse Glimmer – open weights 30B local coding model”
- Comment from 27 Aug 2026 in “Gemini Omni 1.1 Flash”
- Comment from 11 Aug 2026 in “Antirez/h3.c: MiniMax H3 inference engine for Mac computers”
- Comment from 12 Sep 2026 in “Nvidia is the central bank of AI”
Impressive output quality
“it is knowledgeable enough that it can handle a fair amount of anatomy looking and behaving reasonably”
Comment from 11 Aug 2026 in “Antirez/h3.c: MiniMax H3 inference engine for Mac computers” “The model is quite impressive, though obviously it depends a lot on what you're doing.”
Comment from 22 Sep 2026 in “Truman World” “looks better than this at most of the things I've been able to generate locally”
Comment from 27 Oct 2026 in “Genie 4” - Comment from 29 Sep 2026 in “Show HN: Short film made using MiniMax H3 on my home PC”
- Comment from 8 Aug 2026 in “No One Can Afford to Make Myst Games Anymore”
- Comment from 3 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video”
- Comment from 3 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video”
Fast video generation
“Minimax H3 can run exceptionally fast (10 minutes for a 15 second 0.5mp video and that's stock cuda 13)”
Comment from 10 Aug 2026 in “Meta Muse Glimmer – open weights 30B local coding model” “Minimax H3 running local can generate amazing short vids quite fast”
Comment from 10 Aug 2026 in “DeepSeek costs OpenCode Go user $1.14/day; dual DGX breaks even in 24 years” “I could complete within minutes instead of hours and hours.”
Comment from 2 Aug 2026 in “Seedance 2.5” - Comment from 3 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video”
- Comment from 30 Aug 2026 in “Infinite Interdimensional Cable”
Beats rival video models
“MiniMax H3 is already ahead of Seedance 2.0 based on thousands of A/B votes”
Comment from 3 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video” “It's better than all private video models like Veo.”
Comment from 11 Aug 2026 in “Antirez/h3.c: MiniMax H3 inference engine for Mac computers”
What people warn about
Each mark is one comment
Poor, underwhelming results
“the results were quite poor. Unusable for anything, in fact.”
Comment from 3 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video” “H3 results are underwhelming compared to LTX 2.3 so far”
Comment from 2 Aug 2026 in “Seedance 2.5” “H3 does not look or perform at that level”
Comment from 3 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video” - Comment from 1 Aug 2026 in “What Is MiniMax H3? Everything You Need to Know About the Hailuo 3.0 Video Model”
Behind rival video models
“This is still about a year and a half behind Seedance 2.0/ Seedance 2.5”
Comment from 3 Aug 2026 in “MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video” “H3 results are underwhelming compared to LTX 2.3 so far”
Comment from 2 Aug 2026 in “Seedance 2.5”
Other AI video generators people recommend
The four most recommended other AI video generators, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| Seedance 2.0 | 68 | 21 | 76% | 10 | — | Approved |
| MiniMax H3This page | 29 | 11 | 73% | 3 | — | — |
| Google Veo | 20 | 11 | 65% | 8 | — | — |
| Gemini Omni | 16 | 21 | 43% | 2 | — | — |
| Wan video model | 15 | 4 | 79% | 7 | — | Approved |
What people said
40 of the 46 comments that mention it, newest first. Each links to the original comment.
“looks better than this at most of the things I've been able to generate locally”
“Same like Minimax H3 or Qwen3.8-27B”
“insane in the sense that it would not make sense at all”
“MiniMax does produce crisp imagery if you use a high number of steps (20+)”
“The model is quite impressive, though obviously it depends a lot on what you're doing.”
“was a big step forward in the open-weight world though with its ability to do ref2vid”
“works pretty great and you can run it on a 3090”
“Yes its very expensive BUT it's fun”
“which can produce video faster than playback time”
“The int8 release is very capable and runs acceptably on consumer gpus”
“it's pretty good for some quick funny clips”
“I'm happily using Minimax H3 locally on my 12Gb 4070RTX”
“Minimax quality for video generation had been significantly downgraded”
“it is knowledgeable enough that it can handle a fair amount of anatomy looking and behaving reasonably”
“It's better than all private video models like Veo.”
“If you want a good balance of size, speed and quality you should use the int8_convrot model”
“I've been fitting this within 96GB VRAM without issues”
“generating a 15s 480p video with MiniMax H3 in ComfyUI takes an hour and a half”
“Minimax H3 can run exceptionally fast (10 minutes for a 15 second 0.5mp video and that's stock cuda 13)”
“Minimax H3 running local can generate amazing short vids quite fast”
“It's not as good, but it's good enough and it's completely open.”
“clear the model can do amazing stuff, including game scenes”
“Least-censored model of any type I've ever seen from a major lab.”
“download the new minimax h3 model and render a test movie with it”
“These tools are 80-90% of my day now.”
“It is coming up with Seinfeld episodes even with voice overs.”
“H3 does not look or perform at that level”
“MiniMax H3 is already ahead of Seedance 2.0 based on thousands of A/B votes”
“even the demo prompts on the ComfyUI page aren't adhered to by the model”
“the results were quite poor. Unusable for anything, in fact.”
“This is still about a year and a half behind Seedance 2.0/ Seedance 2.5”
“1 minute for a second of footage, that is awesome!”
“impressive. On the other hands aesthetically it all looks painfully bland and generic.”
“the results are spectacular”
“The mouse render is surprisingly good.”
“I could complete within minutes instead of hours and hours.”
“H3 results are underwhelming compared to LTX 2.3 so far”
“it's fun to train it for whatever you need”
“I'd honestly take the slight quality hit for more control and lower costs.”
“Low quality ai slop it feels like.”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.