Update model card

README.md

@@ -64,18 +64,18 @@ pipe = T2AudRefCondPipeline(
 
 refs = ["examples/references/reference_1.wav", "examples/references/reference_2.wav"]
 
-# Game-show buzzer (~8.6 s)
+# Game-show buzzer
 pipe(
     prompt='The speaker from reference 1 fires: "Capital of France?" The speaker from reference 2 instantly: "Paris!" The speaker from reference 1: "Square root of eighty-one?" The speaker from reference 2: "Nine!" The speaker from reference 1: "Year of moon landing?" The speaker from reference 2: "Sixty-nine!" A loud, harsh game-show buzzer rings out — a sustained electronic BZZZZT.',
     ref_audio_paths=refs,
-    duration=8.65,
+    duration=8.5,
 ).save("buzzer.wav")
 
-# Farm at sunrise (~6.4 s) — same two voices, a different scene
+# Farm at sunrise — same two voices, a different scene
 pipe(
     prompt='A farm at sunrise: a rooster crows. Chickens cluck softly throughout. The speaker from reference 1 says with a yawn: "Way too early for this." The speaker from reference 2 chuckles: "Welcome to country life." The rooster crows again.',
     ref_audio_paths=refs,
-    duration=6.41,
+    duration=6.5,
 ).save("farm.wav")
 ```
 
@@ -84,8 +84,9 @@ pipe(
   order of `ref_audio_paths`.
 - Put the spoken words **in quotes**; describe sound effects / ambience in plain prose
   (e.g. *"a loud, harsh game-show buzzer rings out"*, *"a roaring stadium crowd cheers"*).
-- Reference clips should be a few seconds of clean single-speaker speech (any sample rate; mono or stereo).
-- 60 inference steps, guidance ~7, and durations up to ~20 s work best (the model was trained on ≤20 s scenes).
+- Reference clips: clean single-speaker speech, up to ~20 s each — the more the better (any sample rate; mono or stereo).
+- You don't have to follow a strict turn-taking pattern; see the [demo page](https://finmickey.github.io/scena/) for more varied examples.
+- 60 inference steps, guidance ~7, and output durations up to ~20 s work best (the model was trained on ≤20 s scenes).
 
 ## Citation