Skip to content

Open-source AI podcast clipper. Generate vertical clips with face tracking and burned-in captions. CLI, MCP server, web app.

Keep first-run setup off stdout, and find gallery tiles with any ffmpeg build (#238)

Nika Siradzed38a412a5074
diff --git a/backend/services/multicam.py b/backend/services/multicam.py
index 655dcc9..a15f21e 100644
--- a/backend/services/multicam.py
+++ b/backend/services/multicam.py
@@ -443,7 +443,7 @@ def detect_panes(source: Source, samples: int = 8) -> list[list[float]]:
other = 1 if axis == 2 else 2
# App chrome (gutters, margins) is one solid colour; even a plain wall in a
# downscaled, compressed camera picture varies more than this.
- flat = block.std(axis=other).max(axis=0) < 2
+ flat = block.std(axis=other).max(axis=0) < 3
diff = np.abs(np.diff(block, axis=axis)).mean(axis=other).min(axis=0) # a hard edge in every frame
edge = np.concatenate([diff > 40, [False]])
runs, k = [], 0
diff --git a/cli/main.go b/cli/main.go
index e402263..b615dfb 100644
--- a/cli/main.go
+++ b/cli/main.go
@@ -115,7 +115,13 @@ func refreshStudioBundles() {
func ensureRuntime() error {
if _, ok := engine.BackendRoot(); !ok {
fmt.Fprintln(os.Stderr, "First run - setting up podcli (one-time download)...")
- if setup(nil) != 0 {
+ // setup reports on stdout, which belongs to the command being run here:
+ // `--json` callers parse it as exactly one JSON object.
+ stdout := os.Stdout
+ os.Stdout = os.Stderr
+ ok := setup(nil) == 0
+ os.Stdout = stdout
+ if !ok {
return fmt.Errorf("first-run setup failed (see errors above) - run `podcli setup` to retry")
}
} else if err := refreshBackend(); err != nil {