Stop rebuilding your character for every scene. One API handles it all.
Motion transfer for any character, any template video.
Image generation up to 4K.
Portrait + text → lip-synced video.
text-to-video, image-to-video, video-to-video, video editing
All outputs generated via Synts API. No post-processing.
Primitives
Raw model access is table stakes. Variables, jobs, refunds, chaining. Everything you'd otherwise stitch together yourself.
Save any asset by slug. Reference @character1 in any prompt, any session, any model. Same face, guaranteed.
Every ByteDance model through POST /api/generate.
One credit balance. One auth header. One SDK.
vs stitching 5+ vendor APIs
Every output returns a reusable upload_id.
Feed Seedream into Seedance into OmniHuman. Zero re-uploading. Zero URL handling.
Agents can't hallucinate IDs that don't exist
Full PENDING → PROCESSING → COMPLETED state machine. Filter, cancel, retry, webhook.
Cancel = 100% credit refund, auto
Bundle prompt + model + resolution + references into one preset_id.
Reproducible outputs on every deploy.
Version-control your generations
Check exact credit cost before every call: GET /billing/credits/costs.
No surprise bills at any scale.
1 credit = $0.001 · always
Upload any asset once and get a CDN-backed uploadId ready for every model call. No S3 plumbing to build yourself.
GET /api/storage/uploads
Every generation saved automatically. Filter by model, re-download, replay a prompt, or cancel any in-flight job.
GET /api/generation/history
For coding agents
Stable upload_ids on every output. Your agents can't hallucinate
URLs that don't exist. Variables persist across sessions via @mentions.
Zero-config MCP server for Claude Code, Cursor, Windsurf.
Why Synts
We benchmarked the major AI video and image platforms against the four things developers actually care about. The gap is not subtle.
| Capability |
API + Agent layer
|
Other AI platforms
Industry pattern
|
|---|---|---|
| Image generation speed | 10–30 seconds | 30 seconds to 5 minutes |
| Video generation speed | 30 seconds to 3 minutes | 4 to 40+ minutes |
|
Automatic refund on failed jobs
credits return the moment a job errors
|
✓ | ✗ |
|
Refund on disliked output
flag a generation, get credits back
|
✓ | ✗ |
|
Real REST API + SDKs
on every paid plan
|
✓
TS · Python · curl
|
✗
gated, hidden, or none
|
|
Official ByteDance partner
the inference behind TikTok and CapCut
|
✓ | ✗ |
|
Persistent character variables
slug-based, work in every model and session
|
✓ | ✗ |
| Starts at | $19/mo | $7–$25/mo |
Generation queues balloon during peak hours. Paid plans get throttled after launch novelty wears off. A single video can sit pending for 15, 40, or sometimes a full day before timing out.
Failed generations consume credits anyway. Some platforms admit the video "did not generate" but cite "compute costs" to deny refunds. One failed test video can burn the majority of a monthly allowance.
Synts runs on BytePlus, the same inference layer that powers TikTok, CapCut, and Doubao. That means enterprise-grade speed, uptime, and model quality, with a direct partnership behind every call. No wrapping, no aggregating, no middleman.
Most consumer-facing AI video tools have no public API at all. Some gate it to the highest plan ($60+/month), some require enterprise sales conversations, and others route through third parties. Building agents on these stacks is brittle.
Synts gives every credit back. Every time.
Failed generation? Auto-refunded. Don't like the output? One-click refund. The same ByteDance models others wrap, with infrastructure that actually finishes the job.
A free Skool community where we teach you how to use Synts, every ByteDance model, and the broader AI stack you need to ship. Tutorials, live calls, and a feed where you post work and get answers.
Real humans. Fast replies. No bots in the loop.
One credit. One balance. Every model. No hidden costs.
500 free credits. No card. Key in under 60 seconds.