ByteDance
Used in the mock's 'Talking head' workflow (latentsync_v1). 14 credits per clip, about Minutes per clip per run. Sample price.
| LatentSync spec | |
|---|---|
| Status | Beta |
| Type | Lip sync |
| Version | 1.6 |
| Durations | no limit stated |
| Resolutions | 512px face crop |
| Audio | No audio track |
| Licence | Code Apache-2.0; weights OpenRAIL++ |
| Where it runs | Milvey GPUs. The run queues if the pool is full. |
| Typical time | Minutes per clip |
| Price | 14 cr per clip |
Prices and times are samples.
Open models run on Milvey GPUs, and the price follows that GPU time. If the pool is full, the run waits in a queue until capacity returns. It never moves to a paid API. If the run can’t be delivered, the hold is released and you pay 0 cr.
How failed runs are chargedSamples on this site are the makers’ own published outputs, each credited with its source. ByteDance has none listed here so far. Outputs made on Milvey will be added as the beta runs.
Commercial allowed with RAIL use restrictions.
Conditions are summarised from the licence. They are not legal advice, so check the source before client work.
Nothing flagged for this model.
Sources for specs and licence
Join the list. I’ll email you when signups open. No card needed.
Email when signups open · Unsubscribe any time