Seed ASR 2.0
Seed ASR 2.0 has an audio-file API for batch transcription. The current guide accepts files shorter than five hours and smaller than 512 MB.
Technical guide
The optimized streaming endpoint supports two-pass recognition. It can display an initial result, then revise a completed segment through a second recognition pass. Finalized utterances are marked separately.Seed ASR streaming API
Practical use
Practical guidance. Store provisional and final transcripts separately in the interface. Search indexes and subtitles should normally use the final result, or an early recognition error can remain after it was corrected.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | volc.seedasr.auc |
| File resource | volc.seedasr.auc.Seed ASR audio-file API |
Related models
- Seed 2.0 Lite
ByteDance · Seed 2.0
Multimodal model that combines native audio understanding with image, video and text input.
Launch date not confirmedAPI - Seed ASR 1.0
ByteDance · Seed Speech
First-generation Seed speech recognizer with audio-file and legacy bidirectional streaming interfaces.
Launch date not confirmedAPI