AI Inference for Asia.Frontier Quality, Wholesale Pricing.

OpenAI-compatible API · No card to start · From first call to finished cut

Bring your software

One API, every model

Hit a single OpenAI-compatible endpoint and route to the video, image and text models you can't easily reach directly, at low token cost.

# one endpoint, weighted to video-gen
POST /v1/video/generations
{ "model": "doubao-seedance-2-0-260128" }

Explore the models →

Use ours

The VMax video studio

A script-to-screen studio for video editing. Parse a script into cast, scenes and a storyboard, generate every shot, and export. You're billed only for the tokens your renders consume.

Your studio workspace

studio.tokenmax.ai · ready

instant setup

Script parsing Storyboard Preview

Explore the studio →

VMax — Everything to take a script to a finished video

A complete video-editing pipeline. Upload a script and it's parsed into an outline, cast and storyboard. Generate every shot and preview the full cut, with team workspaces, shared asset libraries and usage controls built in.

NOW LIVE!

Seedance 2.5

ByteDance's new-generation professional multimodal video model, now live on VMax. Long narrative, strong reference, precise editing and multilingual — four upgrades in one release.

Try Seedance 2.5 now →

Popular

Seedance 2.0

The production workhorse. Fast, consistent characters and scenes for high-volume short-form — the popular Seedance route on VMax.

Try Seedance 2.0 now →

New · Beta

Wan 3.0

Tongyi Wanxiang's new all-in-one video model. Native 30-second single takes, up to 20 reference assets, and documents or web pages turned straight into video. Now on VMax.

Try Wan 3.0 now →

Start your video-making journey →Try other models →

/studio/storyboard.htmlstoryboard · beta

Storyboard editor Beta

Shot list, per-shot generates, then play in order

API key sk-… optional override

BoardThe Heir Returns

Add shot

Cold open01 · idle · 5s

The confrontation02 · idle · 5s

Rooftop reveal03 · idle · 5s

Cliffhanger04 · idle · 5s

Play all

shot 01 · 5s · dreamina-seedance-2-5-260628

Take 1Take 2

1 · Board

Add shots on the left. List order is the play order.

2 · Shot

Write a prompt, pick a model and duration.

3 · Generate

Generate one shot at a time. Extra runs are takes.

4 · Play

Play all runs selected takes in sequence. No export.

The production pipeline

Script parsing

Upload a script (doc, docx, txt or pdf, up to 20MB and 100k characters). VMax parses it into a structured synopsis, an episode outline, the cast and IP, and a shot-by-shot storyboard script, ready to set up and generate.

the-heir-returns.pdf2.4 MB · 12 episodes parsed

Synopsis2 paragraphs

Episode outline12 scenes

Characters & IP8 roles

Storyboard script86 shots

Cast, scenes & props

Characters, scenes and props are pulled from the script automatically. Build each one three ways: AI-generate, pick from the asset library, or upload your own, in realistic, anime and film styles plus authorized real-person likenesses.

Storyboard generation

Generate every shot per episode, one at a time or in batch. Seedance 2.0 reference frames keep characters and scenes consistent, multimodal references guide each shot, and candidate previews let you pick the take.

00:0000:0500:1100:1600:21

batch Seedance 2.0 ref multimodal

Full-cut preview

Preview the whole video end to end before you ship, with every shot, transition and voice track stitched into one continuous cut.

full cut · 00:00–04:12

Workspaces, teams & usage

Projects

Start a new video from a script, then rename, duplicate or delete it. Switch between everything you've made and the projects shared across your team.

The Heir Returns12 ep · created by me

Midnight Vow8 ep · shared

City of Glass20 ep · shared

Tea House6 ep · created by me

Personal & team libraries

Keep private characters and scenes to yourself, or share a transparent team library of characters, props and scenes that every member can pull into a project. Upload local assets to either.

Personal assetsprivate

+24

Team assetsshared

+61

Teams & sub-accounts

Main accounts create sub-accounts and teams, grant login access and assign members, then set per-member permissions and credit allowances.

MemberRoleStatus

Lin Weilin@studioAdminactive

Mei Chenmei@studioEditoractive

Hao Yuhao@studioViewerinvited

Usage & credits

Track the total credit balance across the main account and every sub-account, see your entitlement packages, and break down exactly where credits are being spent.

72%used

Total balance12,480

Sub-accounts4,210

Spent this cycle32,090

Ph.D.-Led Engineering, Enterprise-Grade AI Inference

Led by leaders and Ph.Ds from UPenn, HKUST, Jardines, and UBS — bringing institutional reliability and carrier-grade scale to APAC's premier token network.

API — Reliable, Secure, Affordable Tokens

One OpenAI-compatible API into Asia's best models, billed per token. Talk to us for access, top up credits, and route your first request in minutes.

Talk to us

Access is invite-only, with no self-serve signup. Tell us what you're building and we'll set you up.

Buy credits

Credits can be used with any model or provider.

Apr 1$XXX

Mar 30$XXX

Get your API key

Create an API key and start making requests. Fully OpenAI compatible.

TOKENMAX_API_KEY

•••••••••••••••••

Video generation

Primary

The core of the platform. Text-to-video and image-to-video for short-form, plus restyle and translation of existing footage inside the studio.

doubao-seedance-2-5-260628Newest

The latest Seedance generation, with the strongest motion coherence and prompt adherence in the family.

text→videoimage→videovoiced

doubao-seedance-2-0-260128Recommended

Flagship text→video and image→video with synced audio. The default route for new video projects.

text→videoimage→videovoiced

doubao-seedance-2-0-fast-260128fast-25%

Lower-latency Seedance 2.0 for high-volume short-form where speed beats polish.

text→videobatch

doubao-seedance-2-0-mini-260615cost-efficient-60%

The cheapest Seedance tier for drafts and high-throughput iteration.

text→videoimage→video

happyhorse-1.1-t2vtext→video

Stylized text-to-video up to 1080P for richer, expressive motion.

text→video1080P

happyhorse-1.1-i2vimage→video

Animate a still into coherent 1080P motion with the HappyHorse family.

image→video1080P

happyhorse-1.1-r2vreference→video

Reference-to-video and edit for consistent characters and restyles.

reference→videoedit

wan2.7-i2vimage→video-12%

Latest Wan image-to-video with strong motion and edit support.

image→videoedit

wan2.7-t2vtext→video-12%

Latest Wan text-to-video, with sharper motion and better prompt following than 2.6.

text→video1080p

wan2.7-r2vreference→video-12%

Reference-to-video on Wan 2.7 for consistent characters and props across shots.

reference→videoedit

wan2.6-t2vtext→video-12%

Cost-efficient text-to-video for high-volume short-form.

text→video

wan2.6-i2vimage→video-12%

Cheapest per-second image-to-video route for bulk generation.

image→videobatch

kling-v3-omni-t2vtext→video

Kling v3 Omni text-to-video for cinematic, high-motion shots.

text→videoomni

kling-v3-omni-i2vimage→video

Kling v3 Omni image-to-video with strong subject consistency.

image→videoomni

kling-v3-t2vtext→video

Standard Kling v3 text-to-video, in std and pro modes for 720p or 1080p output.

text→videostd / pro

kling-v3-i2vimage→video

Standard Kling v3 image-to-video for animating stills with cinematic camera moves.

image→videostd / pro

Image generation

Stills, key-frames and reference assets that feed straight into the shot editor, or stand alone for thumbnails and posters.

doubao-seedream-5-0-pro-260628high fidelity

Flagship Seedream stills with top-tier detail for hero key-frames.

text→imagehi-fi

doubao-seedream-5-0-260128photoreal

Photoreal Seedream generation for posters, thumbnails and assets.

text→image

qwen-image-2.0-prohigh fidelity-12%

The higher-fidelity Qwen tier for hero stills and detailed edits.

text→imagehi-fi

qwen-image-2.0photoreal-12%

Photoreal stills with reliable text-in-image, strong for ad key-frames and posters.

text→imagetext-in-image

wan2.7-image-prohigh fidelity-12%

The pro Wan 2.7 image tier for detailed stills and reference-guided edits.

text→imageimage→image

wan2.6-t2icost-efficient-12%

The cheapest image route for high-volume asset and key-frame generation.

text→imagebatch

Text & reasoning

The models behind every shot: restructuring scripts, writing prompts and powering the analysis agents inside the studio.

qwen3.8-maxNewest-12%

The newest Qwen Max generation, the strongest Alibaba tier for reasoning and agents.

chattools

qwen3.7-maxfrontier-12%

Qwen's most capable tier for hard reasoning and agentic tasks.

chattools

qwen3.7-max-previewpreview-12%

Early access to the next Qwen Max, latest capabilities first.

chattools

qwen3.7-max-2026-06-08snapshot-12%

Pinned dated snapshot of Qwen 3.7 Max for reproducible output.

chattools

qwen3.7-plusbalanced-12%

Balanced Qwen tier for quality prompting at lower cost.

chattools

qwen3.6-max-previewreasoning-12%

Strong long-context reasoning for scripts and storyboards.

chatlong-context

qwen3.6-27bopen-weight-12%

Compact open-weight Qwen for cost-sensitive, high-volume workloads.

chat

qwen3.6-flashfast-12%

Low-latency Qwen for high-volume drafting and classification.

chatfast

glm-5.2reasoning

Strong long-context reasoning for structured scripts and storyboards.

chatlong-context

glm-5.2-fast-previewfast

Preview of the low-latency GLM tier for high-throughput prompting.

chatfast

kimi-k2.7-codecoding

Long-context Kimi tuned for code generation and repo-scale reasoning.

codelong-context

deepseek-v4-pro-260425reasoning

High-end DeepSeek reasoning for planning and complex analysis.

chattools

deepseek-v4-flash-260425fast

Low-latency general model for high-throughput prompting and drafting.

chatfast

deepseek-v4-flash-0731snapshot

Pinned DeepSeek Flash snapshot for reproducible, version-locked output.

chatfast

doubao-seed-2-1-turbo-260628fast

Fast, capable workhorse for script restructuring and bulk prompting.

chatfast

doubao-seed-evolvingadaptive

Continuously updated Doubao Seed tracking the latest capabilities.

chattools

claude-opus-4-8frontier

Top-tier reasoning for the hardest planning and agent steps in the studio.

chattools

claude-opus-4-7reasoning

Deep reasoning and reliable tool use for complex multi-step work.

chattools

claude-opus-4-6reasoning

Proven Opus generation for demanding analysis and long agent runs.

chattools

claude-sonnet-4-6balanced

Fast, capable all-rounder balancing quality and cost for most tasks.

chattools

gpt-5.6-solfrontier

Frontier general intelligence for the toughest reasoning and coding.

chattools

gpt-5.6-terrafrontier

Frontier GPT tuned for grounded, high-accuracy long-context work.

chattools

gpt-5.5reasoning

Strong reasoning and tool use for planning and agent orchestration.

chattools

gpt-5.4general

Dependable general-purpose GPT for chat, drafting and analysis.

chattools

gpt-5.4-minicost-efficient

Cheap, low-latency GPT for high-throughput prompting and drafts.

chatfast

gemini-3.5-flashfast

Low-latency Gemini with long context for bulk and real-time flows.

chatlong-context

gemini-3.1-pro-previewreasoning

Long-context Gemini Pro for structured analysis and multimodal input.

chatlong-context

Speech & audio

Text-to-speech for voiceover, dubbing and subtitles inside the studio's localize flows.

qwen3-tts-instruct-flashtext→speech-12%

Instruction-following TTS behind the studio's voiceover and dubbing flows.

text→speechmultilingual

qwen3-tts-flashfast-12%

The low-latency TTS tier for bulk voiceover and real-time playback.

text→speechfast

Questions, answered

The API is one OpenAI-compatible endpoint that routes your requests to the best video, image or text model at low token cost, so you can bring your own software. VMax is a full video-editing studio built on top of it: script parsing, cast and scene assets, storyboard generation and full-cut preview. Both run on TokenMax tokens, and tokens are what you're billed for.

No. You can call the OpenAI-compatible API directly for video, image and text generation. VMax is the optional step up when you want a workspace, a guided pipeline and a storyboard editor instead of raw calls.

A full video-editing suite: script parsing into outline, cast and storyboard; AI-generated or library characters, scenes and props; shot-by-shot storyboard generation; and full-cut preview. It also includes projects, personal and team asset libraries, sub-account and team management, and usage statistics. Every render consumes TokenMax tokens at your rate.

Most teams think in outputs (a video, an image, a script), not in model names. Pick a category and TokenMax routes to the right model for it. Video leads, with image and text supporting the production pipeline.

The aggregator API is available broadly. Studio availability depends on regional signing and real-name verification. Hong Kong and Macau are supported, with more markets rolling out. Talk to us about your region and we'll confirm what's live today.