2026-11-11 · Hailuo 02 Standard · 6 seconds · 60
HyperFlow H3 Flash مقطع مقطّر. هذا النموذج هو Hailuo 02 Standard
البطاقة تقول 1344×768 ومن 5.2 إلى 15 ثانية. صفحة التسويق تقول 768p و14B. هذا النموذج هو Hailuo 02 Standard لمدة 6 ثوانٍ.
هذا النموذج هو Hailuo 02 Standard، 6 ثوانٍ، 60 رصيدًا. لا يشغّل أوزان HyperFlow H3 Flash ولا MiniMax H3. النص إلى فيديو لا يرسل دقة. 64 رصيدًا تُبقي 4.
المدة
النسبة
الفيديوهات المُنشأة
لا توجد فيديوهات بعد
املأ النموذج وانقر على «إنشاء الفيديو» لإنشاء أول فيديو لك بالذكاء الاصطناعي.
نصيحة: استخدم أوصافًا مفصّلة للحصول على نتائج أفضل
- المطوّر
- Video Rebirth
- تاريخ الإصدار
- 6 نوفمبر 2026
- المتاح على Nano AI
- Hailuo 02 Standard · من 60 رصيدًا لكل فيديو
- آخر تحديث للصفحة
- 11 نوفمبر 2026
المصدر: huggingface.co
Recorders, not HyperFlow H3 Flash frames
touhotus, CC BY 2.0, 2009-02-07. The panel reads REVOX. Mk2010, CC BY-SA 4.0, 2016-01-10. The panel reads NEC, and a reel reads Norelco and Scotch. No person. Not Video Rebirth frames.
ماذا يطبع HyperFlow H3 Flash وماذا يرسل هذا الزر
HyperFlow H3 Flash هو مقطع مقطّر من MiniMax H3 نشرته Video Rebirth. The model card and the GitHub README, read 2026-11-11, say 5.2 to 15 seconds at 24 fps, 1344×768 by default, and a 32 kHz stereo soundtrack. Sampling is 4, 6, or 8 steps, and 6 is the default. Speed for a 5.2 second clip at 6 steps: 8.7 seconds on 4× B200, 36 seconds on 1× H200, and 92 seconds on 1× RTX 5090. The same pages say 13.7B for fl2va and 13.9B for ref2va. The marketing page says 14B, 6-step, 768p in real time on four B200 GPUs, and a 12.7× speedup on a single B200. Q4_K_M is 9.1 GB on the card. The card also says the weights are not an official MiniMax product and are not licensed in the EU, the UK, South Korea, or the US without prior authorization from MiniMax. The GitHub repository metadata lists Created 2026-11-06. This form is Hailuo 02 Standard, 6 seconds, 60 credits. Text to video does not send a resolution. Image to video sends 768P. It does not run HyperFlow weights, does not run MiniMax H3, and does not add a synced soundtrack. 64 credits cover one clip and leave 4.
Sibling pages: /minimax-h3 and /minimax-h3-max are video and use the same Hailuo 02 Standard form. /minimax-h3-image is the still card and opens Qwen Image 2.1. /p-h3-image is Pruna's pack. /minimax-h3-vbvr is a third-party LoRA. This page does not merge them.
HyperFlow H3 Flash next to this form
- Card and README
- 5.2–15 s · 24 fps · 1344×768 · 32 kHz · steps 4, 6, or 8
- Speed on the card
- 8.7 s on 4× B200 · 36 s on 1× H200 · 92 s on 1× RTX 5090 · a 5.2 s clip at 6 steps
- Parameter lines
- 13.7B fl2va · 13.9B ref2va · marketing page says 14B
- Marketing page
- 768p · real time on four B200 GPUs · 12.7× on one B200 · Q4_K_M 9.1 GB is only on the card
- License line
- not an official MiniMax product · EU, UK, South Korea, and the US need prior authorization
- Repository date
- GitHub metadata lists Created 2026-11-06
- This form
- Hailuo 02 Standard · 6 seconds · 60 credits · text to video sends no resolution · image to video sends 768P
Leave the two HyperFlow H3 Flash pages unsmoothed
Card: 1344×768 and 13.7B
5.2 to 15 seconds, 24 fps, 32 kHz, steps 4, 6, or 8. 13.7B and 13.9B. 8.7 seconds, 36 seconds, and 92 seconds. Q4_K_M is 9.1 GB.
Marketing page: 14B and 768p
14B, 6-step, 768p in real time on four B200 GPUs, and 12.7× on a single B200. That page does not print 5.2 or 13.7B.
This button
Hailuo 02 Standard, 6 seconds, 60 credits. Text to video sends no resolution. Image to video sends 768P. No synced audio.
HyperFlow H3 Flash and this form
Read 2026-11-11. 14B and 13.7B stay in different columns.
| Model card | Marketing page | This form | |
|---|---|---|---|
| Length | 5.2 to 15 seconds | not printed as 5.2 | 6 seconds |
| Frame | 1344×768 | 768p | no resolution on text to video; 768P on image to video |
| Size | 13.7B and 13.9B | 14B | not those weights |
| Steps | 4, 6, or 8 | 6-step | not a step control |
| Price here | local weights, Q4_K_M 9.1 GB | real time on four B200 GPUs, 12.7× | 60 credits |
How to run the clip on this page
- 1
Write an empty room
The lines below do not name a song and do not ask for synced audio.
- 2
Keep Hailuo 02 Standard and 6 seconds
The clip is 60 credits. 64 credits leave 4. This is not 5.2 seconds and not 15 seconds.
- 3
Use a sibling page for a different search
Stills are /minimax-h3-image. The fast video tier is /minimax-h3-max. Pruna's pack is /p-h3-image.
Prompts for the 6-second clip
Original lines for Hailuo 02 Standard. They are not HyperFlow showcase prompts.
Still reels
A tape recorder on a desk, both reels still, room tone only, no hands and no song.
Empty speaker
One loudspeaker on a stand in an empty room, the cone still, no person.
Film core
A film reel lying flat on a table, no projector beam and no hands.
Quiet window
Rain on a window and a bowl of noodles on a counter, no person and no captions.
الأسئلة الشائعة
HyperFlow H3 Flash questions
The card and the marketing page do not use the same size line.
هل يشغّل هذا النموذج HyperFlow H3 Flash؟
لا. هذا النموذج هو Hailuo 02 Standard، 6 ثوانٍ، 60 رصيدًا. لا يحمّل الأوزان المقطّرة ولا يضيف صوتًا متزامنًا.
Why do 768p and 1344×768 both stay?
The card says 1344×768 by default. The marketing page says 768p. This form's text to video does not send a resolution. Image to video sends 768P. None of those is edited into one number.
Is the model 14B or 13.7B?
The marketing page says 14B. The card says 13.7B for fl2va and 13.9B for ref2va. Both stay. This site did not count parameters.
Is the speed real time?
The marketing page says 768p in real time on four B200 GPUs and 12.7× on a single B200. The card says 8.7 seconds, 36 seconds, and 92 seconds for a 5.2 second clip at 6 steps. This page does not time a GPU.
Where are the other H3 pages?
/minimax-h3 and /minimax-h3-max use Hailuo 02 Standard. /minimax-h3-image opens Qwen Image 2.1. /p-h3-image is the 55 GB, 30 GB and 20 GB pack. /minimax-h3-vbvr is the LoRA note.
Is Nano AI part of Video Rebirth?
No. Nano AI is not affiliated with Video Rebirth, HyperFlow, or MiniMax. The card says the weights are not an official MiniMax product.
المزيد من Nano AI
Related pages
Nano AI ليست تابعة لـ Video Rebirth أو HyperFlow أو MiniMax. الأسماء تُستخدم فقط لتسمية النموذج. لا تُعرض شعارات.