Give the header back its one line

Put download within reach in the history

Show remaining GPU budget

Show the throughput figures

Make the second pass a choice

Drop the misleading token warning

Show progress on the lock screen

Keep the page alive through the upload

Only say the app may close once it may

Treat a background result as already saved

Report each push stage separately

Add a push self-test

Warn if a job starts without the token

Poll the job instead of driving it

Show uncertainty and let it be corrected

Show what was corrected

Say what the overlap warning means

Await the token status before reporting it

Bigger settings icon, and a way out of the dialog

Recording becomes a plain text link, clearly optional

Glossary moves out of the way, learning stays visible

Rebuild the flow around the actual use case

Remove smoothing step and change markers

Tap a marked line to reveal what was recognised

Smoothing step in the pipeline, glossary shows frequencies

Glossary tab and one-tap term suggestions

Keep screen awake during analysis and allow resuming

Context field for names and terms

Fix HF token never reaching the backend

Show whether requests carry the HF token

Paste your own transcript for best quality

Language selector, default German

Drop accept filter so iOS shows m4a files from iCloud

Drive analysis step by step with per-section progress

Long recordings: wake lock, chunked recording, live progress

Speaker naming before save, language display

Share and download markdown directly from the result view

Pass HF token for ZeroGPU quota attribution

Fix login sheet: hidden handling, keyboard-safe layout, ?key= link

Sprecher-Analyse PWA