minimax-h3-liteMiniMax H3 Lite runs the same MiniMax H3 multimodal video model as minimax-h3, on a budget channel: 480P at $0.01 per second and 768P at $0.02 per second, any whole length from 1 to 15 seconds, so a 768P 10-second clip is $0.20 and a 480P 5-second draft is $0.05. One endpoint covers the useful modes and picks the right pipeline from what you send: a prompt alone is text-to-video; add up to 9 reference images to lock character, product or style; add up to 3 reference audio clips (with at least one image) to drive voice and lip movement; or send a first and last frame for an interpolated shot. Reference images and audio are free. Two things are deliberately left out to keep the price where it is: reference video (motion transfer) and native 2K — for those, use minimax-h3. Output is landscape 16:9 or portrait 9:16. Typical turnaround is about 1.5 minutes for a 480P 3-second clip, 2.5 minutes for 768P 5 seconds and 8 minutes for 768P 15 seconds, plus queue time. Only successful generations are charged.
768P at $0.02/s versus $0.097/s on minimax-h3 — the same MiniMax H3 weights, so the character consistency and prompt-following carry over; what you give up is 2K and reference video
Whole-second control from a 1-second test to a 15-second shot, billed per second: a 768P 10s clip is $0.20, a 480P 5s draft is $0.05
Prompt alone is text-to-video; add reference images, reference audio or a first + last frame and the request is routed to the matching pipeline — no separate model ids to learn
Up to 9 reference images and 3 reference audio clips at no extra charge — audio drives voice and lip movement when paired with at least one image
No reference video, no 2K, and 16:9 or 9:16 only — stated up front so you can pick minimax-h3 when the job needs more
Leave the references below empty for pure text-to-video — every reference input is optional.
Steer character, style or composition. Reference images are free on Lite.
MP3 / WAV, ≤15MB and 2-15s each. Drives voice and lip movement; needs at least one reference image. Free.
Generated video will appear here
Provide URLs and click Generate
MiniMax H3 Lite — это API категории «генерация видео» от MiniMax. MiniMax H3 Lite runs the same MiniMax H3 multimodal video model as minimax-h3, on a budget channel: 480P at $0.01 per second and 768P at $0.02 per second, any whole length from 1 to 15 seconds, so a 768P 10-second clip is $0.20 and a 480P 5-second draft is $0.05. One endpoint covers the useful modes and picks the right pipeline from what you send: a prompt alone is text-to-video; add up to 9 reference images to lock character, product or style; add up to 3 reference audio clips (with at least one image) to drive voice and lip movement; or send a first and last frame for an interpolated shot. Reference images and audio are free. Two things are deliberately left out to keep the price where it is: reference video (motion transfer) and native 2K — for those, use minimax-h3. Output is landscape 16:9 or portrait 9:16. Typical turnaround is about 1.5 minutes for a 480P 3-second clip, 2.5 minutes for 768P 5 seconds and 8 minutes for 768P 15 seconds, plus queue time. Only successful generations are charged. Через платформу APIMODELS доступ к этой модели идёт по единому API с прозрачной оплатой по факту использования. Текущая цена: 480P · per second: $0.01, 480P · 5s: $0.05, 480P · 10s: $0.1, 480P · 15s: $0.15, 768P · per second: $0.02, 768P · 5s: $0.1, 768P · 10s: $0.2, 768P · 15s: $0.3, Reference images (up to 9): $0, Reference audio (up to 3): $0.
Быстро собрать промо-видео бренда под рекламные кампании и продвижение в соцсетях.
Короткие вертикальные ролики для TikTok, Instagram и YouTube.
Показать возможности продукта и снять обучающие ролики, которые повышают конверсию.
Объяснения к курсам, разборы тем и обучающие видео — недорого и на потоке.
MiniMax H3 Lite доступна через APIMODELS по цене: 480P · per second: $0.01, 480P · 5s: $0.05, 480P · 10s: $0.1, 480P · 15s: $0.15, 768P · per second: $0.02, 768P · 5s: $0.1, 768P · 10s: $0.2, 768P · 15s: $0.3, Reference images (up to 9): $0, Reference audio (up to 3): $0. Оплата идёт по факту использования — вы платите только за то, что сгенерировали.
Зарегистрируйтесь на APIMODELS, получите API-ключ и обращайтесь к нашему единому эндпоинту. Есть подробная документация с примерами на cURL, Python и Node.js.
APIMODELS отдаёт ту же самую модель MiniMax H3 Lite через свою агрегирующую платформу. Мы предоставляем единый интерфейс API, поэтому отдельные аккаунты у каждого поставщика не нужны — один ключ открывает доступ ко всем моделям.
Same MiniMax H3 multimodal video model, different channel and price. Lite runs on a budget channel and bills per second: 480P at $0.01/s and 768P at $0.02/s, versus $0.097/s (768P) and $0.145/s (native 2K) on minimax-h3. The model-level strengths — character consistency, prompt following — carry over. What Lite leaves out is native 2K and reference video (motion transfer). Use Lite for drafts, social clips and volume; use minimax-h3 when you need 2K or want to reproduce the camera move of a reference video.
Per second, in two resolution tiers: 480P at $0.01/s (5s $0.05, 10s $0.10, 15s $0.15) and 768P at $0.02/s (5s $0.10, 10s $0.20, 15s $0.30). Any whole length from 1 to 15 seconds, billed for the seconds you ask for. Reference images (up to 9) and reference audio (up to 3 clips) cost nothing extra. No minimum top-up, no subscription, and only successful requests are charged — failures are refunded.
Use the unified video endpoint POST /api/v1/video/generations with model set to minimax-h3-lite. prompt is required; duration is any whole second from 1 to 15 (default 5), resolution is 480p or 768p (default 768p), ratio is 16:9 or 9:16 (default 16:9). Attach references as needed: images (up to 9), audio_list (up to 3, needs at least one image), or first_frame_url + last_frame_url for a first/last-frame shot — each accepts a public URL or base64. The call returns a taskId; poll GET /api/v1/video/generations?task_id=xxx or pass callback_url. Field names match minimax-h3, so switching between the two means changing model only.
No reference video (sending one returns a clear 400 at create time rather than being dropped silently); resolution tops out at 768P — no 1080P or 2K; aspect ratio is landscape 16:9 or portrait 9:16 only; first/last-frame mode needs both frames and cannot be mixed with reference images or audio; reference audio needs at least one reference image. Measured turnaround: about 1.5 minutes for 480P 3s, 2.5 minutes for 768P 5s and 8 minutes for 768P 15s, plus queue time. Result files are kept for 7 days.
В APIMODELS MiniMax H3 Lite соседствует более чем с 60 моделями под одним API-ключом и общим балансом, поэтому выбор — это вопрос соответствия задаче, а не привязки к поставщику. Модель поддерживает 480P / 768P, Multimodal Reference, 1-15s, Up to 9 Ref Images, Up to 3 Ref Audios, First + Last Frame, From $0.01/s; вы можете сравнить её по цене и возможностям с другими моделями категории «генерация видео» и переключиться, изменив одну строку с именем модели — без нового аккаунта и без переписывания интеграции. Все варианты категории «генерация видео» с актуальными ценами — на apimodels.app/models.
MiniMax H3 Lite поддерживает: 480P / 768P, Multimodal Reference, 1-15s, Up to 9 Ref Images, Up to 3 Ref Audios, First + Last Frame, From $0.01/s. Полный список параметров и примеры вызовов — в документации APIMODELS.
Да. APIMODELS отдаёт MiniMax H3 Lite через единый API и один ключ — отдельные аккаунты у каждого поставщика не нужны, и вам не приходится самостоятельно решать вопросы регионального сетевого доступа к каждому из них.
Принимаем Stripe (Visa, Mastercard и другие международные карты) и Alipay. Баланс становится доступен сразу после оплаты.