Fire And Smoke Video Models
Generated on PixelDojo. Produced by PixelDojo's generation pipeline.
WAN 2.7 Text-to-Video is the best model for fire and smoke of the three we tested, because it was the only one that came back at 30 frames per second and 1280 by 720. Fire is fast and bright and smoke is fine moving detail, so frames and pixels decide this more than anything else. We sent one identical forge flare prompt to three models on August 26, 2026, then probed every file. WAN 2.7 Text-to-Video cost 5 credits for 2 seconds at 720p. Vidu Q3 cost 3 credits for 3 seconds at 540p and returned 960 by 528 at 24 frames per second. Grok Video cost 4 credits for 2 seconds at 480p and returned 848 by 480 at 24 frames per second. All three were ready at our first status check, between 222 and 233 seconds after the request. The ranking basis is pixels per frame and frames per second first, then what a delivered second of footage costs.
Side-by-side, same prompt
Every model below ran the identical prompt on PixelDojo so the outputs are directly comparable: “A blacksmith's forge flares up in a burst of orange flame with thick smoke rolling toward the ceiling, embers drifting through the dark workshop, slow close shot”
WAN 2.7 Text-to-Video
233sThe only 30 frames per second file in the test, at 1280 by 720, and the densest at 0.76 megabytes per second of video. 2.5 credits per second at 720p, so a 5 second flame shot is 12.5 credits.
Vidu Q3
222s3 credits for 3 seconds at the 540p tier, which is 1 credit per second, the cheapest rate here. Returned 960 by 528 at 24 frames per second with no audio stream at all.
Grok Video
227s848 by 480 at 24 frames per second, and the loudest audio in the test at a mean of minus 9.7 decibels. Charges 2 credits per second whatever resolution you ask for.
Placed by pixels per frame and frames per second
1. WAN 2.7 Text-to-Video, the only file above 24 frames per second
Our clip cost 5 credits, a 2 second run at the 720p rate of 2.5 credits per second, and it was ready 233 seconds after the request. The file measured 1280 by 720 at 30 frames per second, 2.02 seconds long and 1,534,376 bytes, which works out at 0.76 megabytes for every second of video. Both of those numbers are the highest in this test and both of them are the ones that matter for fire. A flame front changes shape completely between frames, so six extra frames a second is six extra shapes rather than six copies, and smoke is the kind of soft rolling gradient that a thin bitrate turns into banding. It also carried a stereo track at 44.1 kHz, though at a mean of minus 34.5 decibels it is close to silent. Durations run 2 to 15 seconds and 1080p costs 3 credits per second. Skip it when cost is the deciding number, because at 2.48 credits per delivered second it was the most expensive of the three by a wide margin.
2. Vidu Q3, the cheapest second of flame in the test
This run cost 3 credits, a 3 second clip at the 540p rate of 1 credit per second, and it was ready 222 seconds after the request. The file measured 960 by 528 at 24 frames per second, 3.04 seconds long and 1,816,907 bytes, so 0.60 megabytes per second of video. That is the second densest file here and it came from the cheapest tier, which is the interesting part. The rate card runs 1 credit per second at both 360p and 540p, 2 at 720p and 3 at 1080p, and durations go from 1 to 16 seconds, so a 10 second smoke plume at 540p is 10 credits. Audio is off unless you turn it on, and with it off our file carried no audio stream at all rather than a silent one. Skip it if your flame needs 30 frames per second, because 24 is what this model returns and there is no frame rate control to change that.
3. Grok Video, cheap per second but not per pixel
Our clip cost 4 credits, a 2 second run at a flat 2 credits per second, and it was ready 227 seconds after the request. The file measured 848 by 480 at 24 frames per second, 2.04 seconds long and 1,059,062 bytes, which is 0.52 megabytes per second of video, the thinnest of the three. It also arrived with a second video stream carrying a single still frame as cover art, which is worth knowing if you feed these files into a tool that counts streams. The audio is the standout: a stereo 44.1 kHz track at a mean of minus 9.7 decibels and a peak of minus 0.1, by far the loudest here. One quirk decides how you should use it. The price is duration times 2 credits and nothing else, so 720p costs exactly what 480p costs. We ran ours at 480p and got fewer pixels for the same money. Skip the low tier entirely and skip the model when detail in the smoke is the point.
Three forge clips, one prompt, 12 credits total. One file came back at 30 frames per second and two at 24, and the resolutions ran from 848 by 480 up to 1280 by 720.
Why Choose Pixel Dojo for Fire And Smoke Video Models
Professional-quality results with cutting-edge AI technology
Frame rate is a model property, not a setting
None of the three exposes a frames per second control. One returned 30 and two returned 24, at three different resolutions. On flame and embers that choice is made when you pick the model.
A flame test costs 3 to 5 credits
Vidu Q3 accepts 1 second, WAN 2.7 Text-to-Video accepts 2 and Grok Video accepts 1. Checking whether a model understands your forge before you buy a long take is the cheap part.
Resolution does not always change the price
Grok Video charges 2 credits a second at 480p and at 720p. Vidu Q3 charges the same 1 credit a second at 360p and 540p. Reading the rate card is worth more than reading the model name.
How It Works
How we ran the test, so you can repeat it:
Write one prompt and send it unchanged to all three
Ours named the flame, the smoke direction, the embers and the shot type in that order. Changing a word between models turns a comparison into two separate generations.
Buy the shortest duration each model sells for the first attempt
We used 2, 3 and 2 seconds. That was enough to see whether the flare happens at all, which is the part most likely to go wrong, and it kept the whole test at 12 credits.
Probe the files you get back
The catalog listing does not tell you the frame rate. We found 30 frames per second on one file and 24 on the other two, at three different resolutions, and that is what decided the order.
Loved by creators on PixelDojo
Real feedback from people using PixelDojo, pulled from our in-product surveys.
THIS IS SO DOPE !
I have already recommended it to friends
All the tools, plus the guidance
Excellent tools. Ease of use. Well thought out interface. Wide variety of AI tools and features which are up to date with more added each month
I really like how the site is kept up to date, and the leaderboard (where I'm currently #11 most popular artist)
This is the best image generator I've used so far. Well done!
Explore more AI tools on PixelDojo
AI Tools
Compare & Switch
- Best AI Image Generators
- Best AI Video Generators
- Midjourney Alternatives
- Civitai Alternatives
- Runway Alternatives
- Leonardo Alternatives
- Pika Alternatives
- Luma Alternatives
- Magnific Alternatives
- Veo Alternatives
- Flux Alternatives
- Freepik Alternatives
- Seedance Alternatives
- Seedream Alternatives
- Pixverse Alternatives
- GPT Image Alternatives
- Synthesia Alternatives
- Playground Alternatives
- NightCafe Alternatives
- Canva AI Alternatives
- ElevenLabs Alternatives
- ComfyUI Alternatives
- Fal Alternatives
- Replicate Alternatives
Common Questions
Everything you need to know about Fire And Smoke Video Models
Which AI video model is best for fire and smoke?
WAN 2.7 Text-to-Video in our August 26, 2026 test. It was the only one of the three that returned 30 frames per second, it was the only one at 1280 by 720, and it carried the most data per second of video at 0.76 megabytes. Fire changes shape every frame, so extra frames and extra bitrate both show.
What did each fire clip cost?
WAN 2.7 Text-to-Video was 5 credits for 2 seconds at 720p, a rate of 2.5 credits per second. Vidu Q3 was 3 credits for 3 seconds at 540p, a rate of 1 credit per second. Grok Video was 4 credits for 2 seconds at a flat 2 credits per second. The three together came to 12 credits.
How long did the clips take to come back?
All three were finished by our first status check, which landed 222 to 233 seconds after we sent the prompts. That is close enough between them that we did not use turnaround to rank anything on this page.
Can I get fire and smoke at 1080p?
Yes. WAN 2.7 Text-to-Video charges 3 credits per second at 1080p and Vidu Q3 charges 3 credits per second at its 1080p tier. There is also a separate video upscaler at 5 credits if you would rather finish a cheaper render afterwards.
Do these clips come with sound?
Two of the three did. Grok Video returned a stereo 44.1 kHz track at a mean of minus 9.7 decibels and WAN 2.7 Text-to-Video returned one at minus 34.5, which is faint. Our Vidu Q3 run had audio switched off and came back with no audio stream at all.
How do I run these from the API?
POST to /api/v1/models/wan-2.7-t2v/run, /api/v1/models/vidu-q3/run or /api/v1/models/xai-video/run with your prompt and a duration. One key covers every model in the catalog and the request shape does not change between them.