Swiss-German-Speech-Data---Meeting-Notes.md

+## Meeting April 15, 2026
+
+### Participants
+
+Yixuan Xu, Michael Graber
+
+### Previous Action Items
+
+* [ ] Roberto: Licence agreement, also partially for Regionaljournal & Espresso broadcasts (needs to be commercially usable and openly shareable)
+* [x] FHNW: will deliver the Espresso and Regionaljournal data via huggingface dataset with download token to Yixuan
+
+### Discussion Points
+
+* SFT
+  * voices should be as diverse as possible
+  * license
+  * use qwen: timbre generation model ?
+  * Yixuan shares demo on slack for SG
+*  Currently the performance on audio is not very good yet
+  * Swiss German translation is better now, but still behind image 
+  * worse than image quality
+  * 80B audio tokens vs 850B image tokens
+* SRF has not been used yet, will be done later
+* Licencing 
+
+### New Action Items
+
+* [ ] Poke Imanol about state of license agreement
+* [ ] Propose next steps for SFT data
+* [ ] Yixuan: prioritize SRF data ablation
+
+## Meeting cancelled on April 7, 2026
+
+* vacation
+
 ## Meeting, March 18, 2026
 
 ### Participants