YFarmX logoYFarmXIntelligence for an accelerating worldBriefings
YFarmX logoYFarmX
YFarmX

Technology Innovation Institute

Falcon-ASR

speech recognition with Arabic dialect support

Announced 7 October 20261 min readAudio and VoiceLast updated:

Editorial illustration of Falcon-ASR

Key facts

7 Oct 2026
Announced
1.6B
Parameters
Speech recognition
Task
Public demo
Access

Falcon-ASR converts speech into text, with Arabic and Emirati-dialect coverage alongside four other languages. TII announced the 1.6-billion-parameter model on 7 October 2026.

Falcon-ASR is a speech-recognition model that turns recorded speech into text. The Technology Innovation Institute, or TII, announced it on 7 October 2026, with a particular focus on Arabic and Emirati speech.

It transcribes five languages

The 1.6-billion-parameter model supports Arabic, English, French, Spanish and Portuguese. TII describes timestamp support, allowing a transcript to be related to positions in the recording. The Arabic work includes Emirati dialect coverage, which is useful when speech differs from formal written Arabic.

Input Output
Recorded speech in a supported language A text transcript
Speech with supported timestamp processing Text linked to positions in the recording

A public demo provides the first access route

TII links a public Falcon-ASR demonstration. Its announcement describes API access and native applications as planned. The confirmed release is therefore a model announcement with demo access; the announcement’s hosting on Hugging Face does not by itself establish downloadable weights.

TII’s accuracy comparisons are the maker’s own evaluations. Before using transcripts in a workflow, test recordings with the relevant dialect, background sound, speaker overlap and specialist vocabulary. Proper names and numerical details deserve particular review because a readable transcript can still contain errors.

All Audio →