makaron-cli 0.16.0 → 0.16.2

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "makaron-cli",
-  "version": "0.16.0",
+  "version": "0.16.2",
   "description": "Give Claude Code a creative agent. Pass complete creative requests and source media to Makaron Chat.",
   "author": {
     "name": "Versa AI",

package/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "makaron-cli",
-  "version": "0.16.0",
+  "version": "0.16.2",
   "description": "Give Codex a creative agent. Pass complete creative requests and source media to Makaron Chat.",
   "author": {
     "name": "Versa AI",

package/README.md

@@ -107,7 +107,7 @@ npx makaron-cli chat --project auto --image photo.jpg --json -b "make it cinemat
 npx makaron-cli chat --project auto --image img1.jpg --image img2.jpg --json -b "combine these"
 ```
 
-`chat` routes image and video models automatically, but you may select the Agent LLM with `--agent-model`. Accepted values are `auto`, the base model IDs (`gpt-6-luna`, `gpt-6-sol`, `gpt-5.6-terra`, `gpt-5.6-sol`, `gpt-5.6-luna`, `grok-4.6`, `deepseek-v4-pro`, `deepseek-flash`), and the personal-plan routes (`gpt-6-luna-codex-subscription`, `gpt-6-sol-codex-subscription`, `gpt-5.6-terra-codex-subscription`, `gpt-5.6-sol-codex-subscription`, `gpt-5.6-luna-codex-subscription`, `grok-4.6-grok-subscription`). `auto` resolves to GPT-6 Luna through the Codex subscription for eligible accounts (including admins) and through Azure API otherwise. Base GPT-5.6 and GPT-6 IDs select Azure API and base `grok-4.6` selects OpenRouter API; the suffixed IDs explicitly request the corresponding personal plan where authorized. This flag changes only the reasoning/tool-calling Agent LLM.
+`chat` routes image and video models automatically, but you may select the Agent LLM with `--agent-model`. Accepted values are `auto`, the base model IDs (`gpt-6-luna`, `gpt-6.1-sol`, `gpt-6-sol`, `gpt-5.6-terra`, `gpt-5.6-sol`, `gpt-5.6-luna`, `grok-4.6`, `deepseek-v4-pro`, `deepseek-flash`), and the personal-plan routes (`gpt-6-luna-codex-subscription`, `gpt-6.1-sol-codex-subscription`, `gpt-6-sol-codex-subscription`, `gpt-5.6-terra-codex-subscription`, `gpt-5.6-sol-codex-subscription`, `gpt-5.6-luna-codex-subscription`, `grok-4.6-grok-subscription`). `auto` resolves to GPT-6 Luna through the Codex subscription for eligible accounts (including admins) and through Azure API otherwise. Base GPT-5.6, GPT-6 and GPT-6.1 IDs select Azure API and base `grok-4.6` selects OpenRouter API; the suffixed IDs explicitly request the corresponding personal plan where authorized. This flag changes only the reasoning/tool-calling Agent LLM.
 
 ```bash
 # Explicit lower-cost Agent LLM for a controlled comparison
@@ -417,8 +417,20 @@ npx makaron-cli video create --script "continue the camera move into the next be
 
 # 4. Check status
 npx makaron-cli video status <taskId>
+
+# Rebuild only 2–5 seconds and publish the complete video into the project.
+npx makaron-cli video retake --video input.mp4 --start 2 --end 5 --prompt "Make the umbrella red" --project <id> --model seedance-2.5 --wait --json
+```
+
+Video segment editing also works directly from natural-language chat, without a GUI selection:
+
+```bash
+npx makaron-cli chat --project <id> "把 @1 的 18–21 秒切成多机位，其余不变"
+npx makaron-cli chat --project <id> "把 @1 的最后三秒改成夜景"
 ```
 
+Chat resolves the source/range, inspects the footage, expands the edit instruction, and delivers the complete video. It asks only when the source or scope remains ambiguous. Precise cuts, subtitles, dubbing and extension also start from chat. Local editing accepts a 0.1–15s range in a source up to 120s; longer scopes use the appropriate whole-video workflow rather than silently shortening the request.
+
 For project/timeline video editing, use:
 
 ```bash

package/bin/makaron.mjs

@@ -32,11 +32,13 @@ const CHAT_AGENT_MODELS = [
   'auto',
   'gpt-6-luna',
   'gpt-6-sol',
+  'gpt-6.1-sol',
   'gpt-5.6-terra',
   'gpt-5.6-sol',
   'gpt-5.6-luna',
   'gpt-6-luna-codex-subscription',
   'gpt-6-sol-codex-subscription',
+  'gpt-6.1-sol-codex-subscription',
   'gpt-5.6-terra-codex-subscription',
   'gpt-5.6-sol-codex-subscription',
   'gpt-5.6-luna-codex-subscription',
@@ -512,9 +514,9 @@ Options:
   --audio <file|url>        Attach a song, beat, or voice reference. MP3/WAV, repeatable.
   --media-manifest <file|-> Import typed image/video media before this run.
   --skill <id|label|name>   Use an installed skill or auto-install a matched marketplace skill.
-  --agent-model <id>        Agent LLM only: auto, gpt-6-luna, gpt-6-sol, gpt-5.6-terra, gpt-5.6-sol,
+  --agent-model <id>        Agent LLM only: auto, gpt-6-luna, gpt-6.1-sol, gpt-6-sol, gpt-5.6-terra, gpt-5.6-sol,
                             gpt-5.6-luna, grok-4.6, deepseek-v4-pro, deepseek-flash, or a
-                            gpt-6-*-codex-subscription, gpt-5.6-*-codex-subscription or
+                            gpt-6.1-sol-codex-subscription, gpt-6-*-codex-subscription, gpt-5.6-*-codex-subscription or
                             grok-4.6-grok-subscription personal-plan route.
   --background, -b          Submit and print a runId.
   --json                    Output structured JSON (includes per-run "usage" credits).
@@ -2079,7 +2081,7 @@ Commands:
 
   edit [--image <file>] "prompt"     Image edit, text-to-image, or transparent PNG
   analyze --video <file|url>         Analyze video content
-  video script|create|status         H3 Max, Wan, Seedance, Grok, lip-sync, and more
+  video script|create|retake|status  H3 Max, Wan, Seedance, Grok, lip-sync; interval retake
   music create|status                Music generation
 
   admin                              Admin commands (skills, credits, upload, set-admin)
@@ -2232,6 +2234,7 @@ Not sure which built-in skill to use? Start with:
   } else if (topic === 'video') {
     if (subtopic === 'script') console.log('Usage: makaron video script --image <file> [--image <file>] [--lang en|zh] "direction"');
     else if (subtopic === 'create') printVideoCreateHelp();
+    else if (subtopic === 'retake') console.log('Usage: makaron video retake --video <url|file> --start <seconds> --end <seconds> --prompt "change" [--model seedance-2.5-eco|seedance-2.5|fal-h3-max] [--project <id>] [--wait] [--json] [--request-id <uuid>] [--audio-mode original|generated]');
     else if (subtopic === 'status') console.log('Usage: makaron video status <taskId> | --snapshot <snapshotId> [--wait]');
     else printVideoHelp();
   } else if (topic === 'music') {
@@ -3085,6 +3088,47 @@ if (!command || command === '--help' || command === '-h' || command === 'help')
     const text = result?.content?.find(c => c.type === 'text')?.text;
     if (text) console.log(text);
 
+  } else if (sub === 'retake') {
+    const params = { model: 'fal-h3-max' };
+    let wait = false, json = false;
+    for (let i = 2; i < args.length; i++) {
+      const flag = args[i];
+      if (flag === '--wait') wait = true;
+      else if (flag === '--json') json = true;
+      else if (flag === '--video' && args[i + 1]) params.video_url = args[++i];
+      else if (flag === '--start' && args[i + 1]) params.start = Number(args[++i]);
+      else if (flag === '--end' && args[i + 1]) params.end = Number(args[++i]);
+      else if (flag === '--audio-mode' && args[i + 1]) params.audio_mode = args[++i];
+      else if (flag === '--prompt' && args[i + 1]) params.prompt = args[++i];
+      else if ((flag === '--model' || flag === '--video-model') && args[i + 1]) params.model = args[++i];
+      else if (flag === '--project' && args[i + 1]) params.project_id = args[++i];
+      else if (flag === '--request-id' && args[i + 1]) params.request_id = args[++i];
+      else { console.error('Usage: makaron video retake --video <url|file> --start <seconds> --end <seconds> --prompt "change" [--model seedance-2.5-eco|seedance-2.5|fal-h3-max] [--project <id>] [--wait] [--json] [--request-id <uuid>] [--audio-mode original|generated]'); process.exit(1); }
+    }
+    if (!params.video_url || !params.prompt || !Number.isFinite(params.start) || !Number.isFinite(params.end)
+      || params.start < 0 || params.end - params.start < .1 || params.end - params.start > 15
+      || (params.audio_mode && !['original','generated'].includes(params.audio_mode))
+      || !['seedance-2.5-eco', 'seedance-2.5', 'fal-h3-max'].includes(params.model)) {
+      console.error('Retake needs one video, a prompt, a valid 0.1–15 second interval, and a supported model.'); process.exit(1);
+    }
+    if (!isHttpUrl(params.video_url)) {
+      if (!params.project_id) { console.error('Local video upload requires --project <id>.'); process.exit(1); }
+      const valid = validateVideoFileForAnalysis(params.video_url);
+      if (!valid.ok) { console.error(valid.error); process.exit(1); }
+      params.video_url = await uploadFileViaSignedUrl(baseUrl, headers, params.project_id, params.video_url, valid.mime);
+      if (!params.video_url) process.exit(1);
+    }
+    const receipt = await callMcpTool(baseUrl, headers, 'makaron_retake_video', params);
+    const raw = receipt?.content?.find(c => c.type === 'text')?.text;
+    let result;
+    try { result = JSON.parse(raw); } catch { throw new Error(raw || 'Retake returned no receipt.'); }
+    if (wait && result.success && result.taskId && !result.videoUrl) {
+      const url = await pollVideo(baseUrl, headers, result.taskId, result.snapshotId);
+      result = { ...result, status: url ? 'completed' : 'processing', videoUrl: url || undefined };
+    }
+    console.log(json ? JSON.stringify(result) : [result.message, result.taskId && `Task: ${result.taskId}`, result.videoUrl && `Video: ${result.videoUrl}`].filter(Boolean).join('\n'));
+    if (!result.success) process.exitCode = 1;
+
   } else if (sub === 'create') {
     let images = [];
     const videos = [];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "makaron-cli",
-  "version": "0.16.0",
+  "version": "0.16.2",
   "description": "Talk to Makaron Agent from the terminal — create projects, edit images, generate videos",
   "type": "module",
   "scripts": {

package/skills/makaron/SKILL.md

@@ -99,7 +99,7 @@ npx makaron-cli chat --project auto --image photo.jpg --json -b "make it cinemat
 npx makaron-cli chat --project auto --image img1.jpg --image img2.jpg --json -b "combine these"
 ```
 
-`chat` routes image and video models automatically. Use `--agent-model` only when the user explicitly asks to select or compare the reasoning/tool-calling Agent LLM. Accepted values are `auto`, the base model IDs (`gpt-6-luna`, `gpt-6-sol`, `gpt-5.6-terra`, `gpt-5.6-sol`, `gpt-5.6-luna`, `grok-4.6`, `deepseek-v4-pro`, `deepseek-flash`), and the personal-plan routes (`gpt-6-luna-codex-subscription`, `gpt-6-sol-codex-subscription`, `gpt-5.6-terra-codex-subscription`, `gpt-5.6-sol-codex-subscription`, `gpt-5.6-luna-codex-subscription`). `auto` uses GPT-6 Luna through the Codex subscription for eligible accounts (including admins) and through Azure API otherwise; base GPT-5.6 and GPT-6 IDs select Azure API, while suffixed IDs explicitly request the personal plan where authorized. Never put an image or video model ID in `--agent-model`.
+`chat` routes image and video models automatically. Use `--agent-model` only when the user explicitly asks to select or compare the reasoning/tool-calling Agent LLM. Accepted values are `auto`, the base model IDs (`gpt-6-luna`, `gpt-6.1-sol`, `gpt-6-sol`, `gpt-5.6-terra`, `gpt-5.6-sol`, `gpt-5.6-luna`, `grok-4.6`, `deepseek-v4-pro`, `deepseek-flash`), and the personal-plan routes (`gpt-6-luna-codex-subscription`, `gpt-6.1-sol-codex-subscription`, `gpt-6-sol-codex-subscription`, `gpt-5.6-terra-codex-subscription`, `gpt-5.6-sol-codex-subscription`, `gpt-5.6-luna-codex-subscription`). `auto` uses GPT-6 Luna through the Codex subscription for eligible accounts (including admins) and through Azure API otherwise; base GPT-5.6, GPT-6 and GPT-6.1 IDs select Azure API, while suffixed IDs explicitly request the personal plan where authorized. Never put an image or video model ID in `--agent-model`.
 
 ```bash
 npx makaron-cli chat --project auto --agent-model deepseek-v4-pro --json -b "make a 20s badminton video"
@@ -341,8 +341,26 @@ npx makaron-cli video create --script "make it warmer and cinematic" --video htt
 
 # 4. Check status
 npx makaron-cli video status <taskId>
+
+# Retake only a selected interval; publish the complete video to a project.
+npx makaron-cli video retake --video input.mp4 --start 2 --end 4 --prompt "Make the book red" --model seedance-2.5 --project <id> --wait --json
+```
+
+Video segment editing also works directly from natural-language chat, without a GUI selection:
+
+```bash
+npx makaron-cli chat --project <id> "把 @1 的 18–21 秒切成多机位，其余不变"
+npx makaron-cli chat --project <id> "把 @1 的最后三秒改成夜景"
 ```
 
+Chat resolves the source/range, inspects the footage, expands the edit instruction, and delivers the complete video. It asks only when the source or scope remains ambiguous. Precise cuts, subtitles, dubbing and extension also start from chat. Local editing accepts a 0.1–15s range in a source up to 120s; longer scopes use the appropriate whole-video workflow rather than silently shortening the request.
+
+Segment edits have two intents: modify changes the inside of the existing sequence and preserves its opening/closing states; replace discards a shot and does not lock its original endpoints. A camera change alone does not mean replacement. A same-request model comparison reuses the original source and interval, not the last generated result.
+
+Chat segment inspection also reads source speech through ASR, reusing cached transcripts. It returns selected speech with source/output timecodes separately from adjacent speech, so the edit can follow the retained narration. Original audio is the default; Chat may select `audio_mode: generated` for requested new sound inside the selection. ASR does not guarantee music-beat or lip synchronization, and unavailable/untimed speech is reported rather than assigned invented timings. Direct `video retake` uses the caller's completed prompt and does not perform Agent interpretation or ASR planning.
+
+`video retake` uses original-source seconds, accepts a 0.1–15s range in a source up to 120s, and automatically replaces only that range while preserving duration. Audio defaults to original; `--audio-mode generated` uses synchronized new audio inside the selection and source audio outside. Models: `fal-h3-max` (default), `seedance-2.5-eco` (preferred Seedance option), and explicitly selected native `seedance-2.5`. For sources shorter than 2 seconds, explicitly use Seedance. Local files require `--project`. Keep the root taskId and request ID; resume polling or replay the same request ID instead of paying for another generation.
+
 `video create` returns a provider task id and does not create or update a Makaron project timeline. For project/timeline video editing, use:
 
 ```bash