@writepanda/mcp 1.217.0 → 1.218.0

package/bin/server.mjs

@@ -4318,7 +4318,7 @@ const TOOLS = [
 	{
 		name: "audio_enhance",
 		description:
-			"Enhance voice: THE audio cleanup step (one call). Runs the studio chain on speech, per source (a podcast's host and each guest separately): noise removal, coughs/bumps out of pauses, mouth-click repair, broadcast tone, boom control, level riding (lifts very quiet speech too), de-esser, quieter pauses, -16 LUFS / -1 dBTP. strength: natural (default), light (gentle tidy), strong (assertive broadcast). keepBackground true for vlogs / ambience where the room is part of the video (no noise removal, event remover or pause expander; level, EQ and loudness still run). echo true only when the user mentions echo / reverb or the room is bare. restore true = Restore voice quality (TF-Restormer) for thin, muffled or call-quality mics only (Apple silicon; slow; never on a good mic). reference = a clip id or file whose voice tone to match. off true restores the original audio. The noise-removal engine is automatic. Every run processes the original media, so running again replaces the result (never stacks). ASYNC: { jobId }, job_wait; the result reports before/after loudness (LUFS, LRA, true peak, speech over pauses, tone deviation) and the engine per source: tell the user.",
+			"Enhance voice: THE audio cleanup step (one call). Runs the studio chain on speech, per source (a podcast's host and each guest separately): noise removal, coughs/bumps out of pauses, mouth-click repair, broadcast tone, boom control, level riding (lifts very quiet speech too), de-esser, quieter pauses, -16 LUFS / -1 dBTP. strength: natural (default), light (gentle tidy), strong (assertive broadcast). keepBackground true for vlogs / ambience where the room is part of the video (no noise removal, event remover or pause expander; level, EQ and loudness still run). Enhance voice already removes room echo with the Studio AI model; only use echo when the result reports the model did not run (fallback) (denoiser deepfilternet3, or keepBackground); when the model runs, echo is skipped and the result says echoSkipped: ai-model-removes-echo (relay echoSkippedReason, do not retry). restore true = Restore voice quality (TF-Restormer) for thin, muffled or call-quality mics only (Apple silicon; slow; never on a good mic). reference = a clip id or file whose voice tone to match. off true restores the original audio. The noise-removal engine is automatic. Every run processes the original media, so running again replaces the result (never stacks). ASYNC: { jobId }, job_wait; the result reports before/after loudness (LUFS, LRA, true peak, speech over pauses, tone deviation) and the engine per source: tell the user.",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -4339,7 +4339,12 @@ const TOOLS = [
 					default: false,
 					description: "Keep the room sound (vlogs, ambience).",
 				},
-				echo: { type: "boolean", default: false, description: "Reduce room echo." },
+				echo: {
+					type: "boolean",
+					default: false,
+					description:
+						"Reduce room echo. Enhance voice already removes room echo with the Studio AI model; only use echo when the result reports the model did not run (fallback). Skipped (echoSkipped) whenever the model runs.",
+				},
 				restore: {
 					type: "boolean",
 					default: false,
@@ -4359,14 +4364,18 @@ const TOOLS = [
 	{
 		name: "audio_clean",
 		description:
-			"Deprecated alias of audio_enhance strength light (noise removal, gentle level and tone, -16 LUFS), echo true adds Reduce echo; clips already enhanced are skipped unless clipId is given. Prefer audio_enhance. ASYNC: { jobId }, job_wait.",
+			"Deprecated alias of audio_enhance strength light (noise removal, gentle level and tone, -16 LUFS), echo true adds Reduce echo only when the Studio AI model does not run (it already removes room echo; the result then says echoSkipped); clips already enhanced are skipped unless clipId is given. Prefer audio_enhance. ASYNC: { jobId }, job_wait.",
 		inputSchema: {
 			type: "object",
 			properties: {
 				id: { type: "string" },
 				path: { type: "string" },
 				clipId: { type: "string" },
-				echo: { type: "boolean", description: "Only when the user mentions echo / reverb." },
+				echo: {
+					type: "boolean",
+					description:
+						"Enhance voice already removes room echo with the Studio AI model; only use echo when the result reports the model did not run (fallback).",
+				},
 				normalize: { type: "boolean", description: "Ignored (quiet speech is always lifted)." },
 			},
 		},
@@ -5717,7 +5726,8 @@ const TOOLS = [
 	},
 	{
 		name: "llm_generate_timestamps",
-		description: "Generate YouTube chapter timestamps from the transcript.",
+		description:
+			"Generate YouTube chapter timestamps for the edited video: edited-timeline times from the words that survive the cuts (first 0:00, each >= 10 s, all before the end); the local LLM only writes labels.",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -6233,7 +6243,7 @@ Before an editing pass, read the skill's "Which tool for which moment" section (
 3. Fix misheard names / brands with transcript_find_replace (transcript_insert_words for dropped words).
 4. transcript_find_issues, then transcript_delete_words on each duplicate take / false start (keep the LAST take; keep severity "low" ones).
 5. transcript_remove_silences.
-6. audio_enhance (Enhance voice) on clips not yet enhanced: strength natural; keepBackground true for vlogs / ambience; echo only when the room is echoey.
+6. audio_enhance (Enhance voice) on clips not yet enhanced: strength natural; keepBackground true for vlogs / ambience; no echo (Enhance voice already removes room echo with the Studio AI model; only use echo when the result reports the model did not run (fallback)).
 7. The style, after steps 2-5 (they shift the timeline). The user's named style, recipe, creator or a saved memory wins. Otherwise call recipe_pick: it returns the recipes that fit this project's format and footage, each with pickWhen, plus defaultId. Read the transcript, choose the ONE candidate whose pickWhen matches what is said (defaultId when none clearly fits), do its "before" step if any, then recipe_apply_style + recipe_render and follow the prompt: the recipe decides the style (captions, pacing, kinds of moves, music, transitions). Graphics inside it are designed by you: a named template is the reference look; author your own (motion_render_html, same palette and type) when a beat needs something the template can't show. (An app without recipe_pick, 2.0.5 and older: recipe_list with format short or long, and choose by description and footage.) Tell the user which recipe and why in one line and name one or two alternatives. Only when no candidate exists or the user asked for a plain clean-up: captions (caption_toggle, template glowStack), a few motion graphics, emphasis zooms.
 8. B-roll, last: pick the moments yourself from what's said (any language), check them with project_check_broll, show the user your moments and ASK: their own images, generated on their image connector (only if one is connected; it uses their credits), or no B-roll. Never generate or invent images without their answer. Place what they give with project_add_broll. Nobody to ask: skip it and say so.
 9. llm_generate_title / description / timestamps, then preview_show and ask before export_start. Never export before the user has seen the preview.

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@writepanda/mcp",
-	"version": "1.217.0",
+	"version": "1.218.0",
 	"description": "Model Context Protocol server for PandaStudio. Exposes the desktop video editor's automation surface to Cursor, Continue, Cline, Claude Desktop, and any MCP-compliant client.",
 	"keywords": [
 		"pandastudio",