- Who it helps
- Developers evaluating multimodal classification and routing.
- Availability
- The model card lists downloadable weights under Apache 2.0 and a Sprag API offer free through October 2026. No hosted Hugging Face inference provider was listed.
Choose the labels explicitly
The card describes a Qwen3-Omni-based model with a decision head: a component that returns scores over supplied choices in one pass. An example evaluation could distinguish “speech”, “music” and “neither”, using clips with known labels.
Keep audio and frames separate
Video frames do not supply the soundtrack; audio must be supplied separately where needed. Returned probabilities also need calibration checks before they are treated as dependable confidence estimates.
What you can do
- Read the input examples and evaluate a small labelled set before relying on the returned scores. Open reference ↗
Where the evidence stops
The model card was read without executing its custom code. Accuracy, calibration, hardware requirements and API terms after October remain unverified.
Sources read. Public sources read October 23, 2026 UTC. No external product workflow, model inference or reported experiment was executed.
Sources & further reading
Hugging Face · implementation
SPEV model card ↗
Selected for the card describes multimodal decision scoring, downloadable weights and a free API offer through October 2026. Custom model code was not executed.
SpragAI · page checked 2026-10-23
Coverage note: Primary documentation was read. Independent workflow execution and original discussion or demonstrations across YouTube, X and Reddit remain incomplete; missing engagement counts are not inferred.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
SPEV scores a supplied set of named choices from multimodal input. It is intended for decisions such as selecting which label fits a clip.