Hailuo 2.3 Fast: the still is the contract (6s, 10s, 28cr)
Hailuo 2.3 Fast is image-to-video. If the label is wrong, every second is waste.
Hailuo 2.3 Fast is image-to-video. If the label is wrong, every second is waste.
The catalog: fast MiniMax Hailuo 2.3 image-to-video — quick, cost-efficient cinematic clips from a single image. Requires an image. Durations: 6s and 10s. Credits: 28. Discounted row (6s listed 28 full / 14 discounted; 10s listed 56 full / 28 discounted). No max-output-resolution chip. Catalog audio is unset — unlike some talking rows, this is not a native-audio promise. Features: Fast, Image-to-video, Cinematic.
Two durations. One still. That is the whole machine.
One image, then 6s or 10s
Hailuo 2.3 Fast does not have a text-to-video sibling on this catalog path. If you only have a sentence, you are in the wrong family. Lock a still on a text-to-image row, crop it to the cut, then come here.
Prompt the motion only. The photo already is the scene. "Slow push-in, steam, she smiles, camera static otherwise." A second competing action in a 6-second clip is how Fast still looks expensive. 10s is for a gesture that lasts 10 seconds, not for a second location.
The AI image to video tool is the door. Hailuo's roster also publishes Hailuo 2.3 Pro (49 credits, 6s / 10s, 1080p, audio false) and Hailuo 2.3 Standard (28 credits, 6s / 10s, 1080p, audio false). Fast is the discounted, cost-efficient cinematic pass from a single image. Pro is a different bill. Do not "upgrade in the prompt." Change the row.
28 credits is the listed Fast rate
The title number is the catalog credits field: 28. Because the row is discounted, the 6s tier can land at 14 and the 10s keeper at 28 on the discounted side of the matrix. That is still a video rate. It is not a 1-credit still. If the still is wrong, both 6s and 10s are waste — cheaper waste than Pro, waste all the same.
There is no duration between 6 and 10. You do not get 8s. Size the action to 6, or commit to 10. Writing a 15-second phrase into this enum is how the model compresses the ending.
The catalog does not list native audio on Fast. Do not write a monologue into the motion prompt and expect a voice file. If the mouth is the shot, pick a talking or lipsync row. If you need VO, generate a mute-looking plate and lay TTS after. MiniMax's own family tip is narrative scene progression — useful for the picture beat, not a substitute for a speech model.
The still is the only contract Fast will honor
Cinematic, fast, cost-efficient: those adjectives still start from image_url. A soft pack label in the jpeg is a soft pack label for 10 seconds of cinema. Approve the still. Then buy 6s. Then, if the motion is the keeper, buy 10s. Do not start on 10s as a way to "see more." You will see more of the same defect.
Best image-to-video is the ranking. Hailuo 2.3 Fast is the two-duration, discounted, still-required member — not a text-to-video sketch model.
FAQ
Can I run Hailuo 2.3 Fast from text only?
No. It requires an image. There is no T2V variant on this Fast card. Generate or shoot the still, then animate.
Why only 6s and 10s?
That is the catalog duration list. Not 4s, not 8s, not 15s. Write one continuous action that fits 6, or one that fits 10.
Does Fast include a soundtrack?
The catalog does not list audio on this row. Treat it as a picture pass. Add voice or music after, or pick a native-audio model when the mouth is in frame.
When should I use Pro instead of Fast?
When you want the Pro card's listed 1080p path and 49-credit rate, not when a prompt adjective says "more cinematic." Fast is the discounted 28-credit, 6s/10s I2V row. Change the model, not the prose.