MiniMax H3 appears on TongAI as several video variants, including Spicy, Max, Max Turbo, Enhanced, Multi-Angle, and compatible LoRA routes. The exact starting input and available resolution depend on the variant, so the model page separates prompt-only, frame-guided, and reference-led uses.
Use MiniMax H3 Text to video · Frame animation · Reference video · Included audio on supported variants
TongAI has no prompt restrictions, including for adult creative themes. Choose the subject, tone, and visual style that fit your idea, and write the direction in your own words.
Model overview
MiniMax H3 is a family of video models with distinct routes for generating from text, animating an image, and using richer reference inputs. TongAI presents the family as multiple named variants rather than flattening their capabilities into one promise.
Max and Turbo routes prioritize different balances of cost, speed, and frame input. Enhanced and reference variants expose other resolution or media options. Some H3 routes include audio in the generated video. Because availability can change, use the live model picker to confirm the variant attached to the workflow you want.
Use a prompt-only H3 route when the scene can be invented from language. Use a frame route when composition and appearance are already designed. Choose a reference-to-video route when several views or compatible video/audio guidance should inform the clip. The model picker exposes only media fields supported by that route.
For a reusable character, use sharp reference images that agree about facial structure, body proportions, age, and wardrobe. Mention the subject consistently and avoid asking for camera moves that hide or radically distort defining features. Compare takes rather than assuming a longer or higher-resolution variant automatically improves identity consistency.
TongAI has no prompt restrictions for MiniMax H3, including for adult creative themes. Paid generation can continue while you have credits, and TongAI displays the selected variant's per-second price before you submit.
Current TongAI pricing
There is no fixed daily generation cap. Each completed generation uses credits at the rate shown for its workflow, model, resolution, and duration.
| Workflow | Variant | Resolution | Starting rate |
|---|---|---|---|
| LoRA video | MiniMax H3 Singularity · image to video · WaveSpeed | Model default | 32 |
| LoRA video | MiniMax H3 Singularity · reference to video · WaveSpeed | Model default | 30 |
| LoRA image | MiniMax H3 · image edit · WaveSpeed | Model default | 5 |
| LoRA video | MiniMax H3 · image to video · WaveSpeed | Model default | 25 |
| LoRA video | MiniMax H3 · reference to video · WaveSpeed | Model default | 30 |
| LoRA image | MiniMax H3 · text to image · WaveSpeed | Model default | 4 |
| LoRA video | MiniMax H3 · text to video · WaveSpeed | Model default | 25 |
| Text to video | Tong Motion Muse | 480p | 8 |
| Video from frames | MiniMax H3 Max Multi-Angle | 480p | 6 |
| Text to video | MiniMax H3 Max | 480p | 2 |
| Reference video | MiniMax H3 Max | 480p | 6 |
| Text to video | MiniMax H3 Max Turbo | 480p | 1 |
| Director video | MiniMax H3 Max Turbo | 480p | 2 |
| Text to video | MiniMax H3 Enhanced | 768p | 10 |
| Reference video | MiniMax H3 Enhanced | 768p | 10 |
| Reference video | MiniMax H3 | 768p | 10 |
| Director video | MiniMax H3 Max Turbo | 480p | 1 |
These are starting rates from TongAI’s public catalog. Higher resolutions and other settings may cost more; Studio shows the final total before submission.
Start from the input you already have, then choose the compatible model route and review its exact price.
Select Text to Video, Frames, Create Video, or LoRA Studio according to whether the request begins with words, a frame, a reusable subject, or an adapter.
Read the model label and route details. Max, Turbo, Enhanced, Multi-Angle, and Spicy variants do not expose identical controls.
Describe what changes during the clip, how the camera moves, and what should be heard. Keep dialogue concise and attribute it to a visible speaker.
Review the per-second credit total, generate, and check video continuity, audio, identity, hands, lip motion, and the final frame.
Prompt ideas
These are starting prompts, not verified example outputs or guarantees of a particular result.
Medium close-up in a quiet diner at night. The subject looks off-camera and says, 'We leave before sunrise.' Low room tone, subtle neon flicker, slow push-in.
Begin from the supplied portrait. A soft wind moves the coat and hair while the camera makes a slow half-circle; preserve the face and muted background.
Use the variant exposed in the workflow matching your input: text, starting frame, multiple references, or LoRA. Then compare resolution, audio, duration, and price instead of choosing by name alone.
Several current H3 routes include audio. The Studio settings and model description show whether audio is included, optional, inherited from reference footage, or unavailable.
Reference-capable variants can improve continuity by using several consistent views. They do not guarantee an identical face or body in every frame, so each output requires review.
Choose the compatible route, review the live credit total, and generate without a fixed daily cap.