Granite Speech 5.0 2B
Granite Speech 5.0 2B is an audio-to-text model built around a Granite 4.0 1B Base language backbone. It recognizes English, French, German, Spanish, Portuguese and Japanese speech.
Model design and capabilities
The model supports translation to and from English for its six main languages, plus English speech to Italian or Mandarin text. It adds language-ID prompting and keyword lists of up to 2,000 entries. The architecture includes a 16-block Conformer encoder and a trained audio projector.Granite Speech 5.0 2B model card
Practical use
Practical guidance. Use keyword guidance for product names or technical terms that ordinary transcription misses. Check whether the hints cause false substitutions in unrelated speech. Score transcription and translation separately; a correct transcript does not prove that the translated meaning is complete.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| Model identifiers | ibm-granite/granite-speech-5.0-2b |
| License | Apache 2.0Granite Speech 5.0 2B model card |
| Parameters | Approximately 2B across the speech-language system.Granite Speech 5.0 2B model card |
Related models
- Granite Speech 4.1 2B
IBM · Granite Speech
A speech-language model for recognition and bidirectional translation between English and five other languages.
Launch date not confirmedAPI · Downloadable · Provider product - Granite Speech 5.0 470M TurboCTC
IBM · Granite Speech
A non-autoregressive speech recognizer that produces capitalized, punctuated text in seven languages.
Launch date not confirmedDownloadable