@crievo/omni 0.0.0-stage → 0.1.1

package/omnivideo.config.example.json

@@ -0,0 +1,437 @@
+{
+  "voice": {
+    "pronunciation": {
+      "Haiku 5.5": "Haiku cinco cinco",
+      "Haiku 4.5": "Haiku quatro cinco"
+    },
+    "order": [
+      "cartesia",
+      "fish",
+      "voicestudio",
+      "elevenlabs"
+    ],
+    "profiles": [],
+    "default_profile": null,
+    "parallel": {
+      "lines_per_worker": 4,
+      "max_workers": 4
+    },
+    "fish": {
+      "key_env": "FISH_API_KEY",
+      "voice_id": "",
+      "model": "s2.1-pro",
+      "fallback_model": "s2.1-pro-free",
+      "speed": 1,
+      "timeout_s": 120
+    },
+    "cartesia": {
+      "key_env": "CARTESIA_API_KEY",
+      "voice_id": "",
+      "model": "sonic-3.6",
+      "language": "pt",
+      "api_version": "2026-08-14",
+      "speed": 1,
+      "timeout_s": 120
+    },
+    "elevenlabs": {
+      "key_env": "ELEVENLABS_API_KEY",
+      "voice_id": "",
+      "model": "eleven_multilingual_v2",
+      "language": "pt",
+      "speed": 1,
+      "stability": 0.5,
+      "similarity": 0.75,
+      "timeout_s": 120
+    }
+  },
+  "usage": {
+    "fish": {
+      "usd_per_million_bytes": 15
+    },
+    "cartesia": {
+      "plan": "pro",
+      "monthly_quota": 100000,
+      "credits_per_char": 1,
+      "admin_key_env": "CARTESIA_ADMIN_API_KEY",
+      "cycle_day": 1
+    }
+  },
+  "voicestudio": {
+    "url": "http://127.0.0.1:3900",
+    "python": "",
+    "repo": "",
+    "profile": "",
+    "language": "Portuguese",
+    "num_step": 16,
+    "speed": 1,
+    "timeout_s": 420
+  },
+  "motion": {
+    "render": "tools/render.mjs",
+    "jobs": 3,
+    "compose": {
+      "parallel": 8
+    },
+    "fps": 30,
+    "formats": {
+      "9:16": {
+        "w": 1080,
+        "h": 1920
+      },
+      "16:9": {
+        "w": 1920,
+        "h": 1080
+      }
+    },
+    "default_format": "9:16",
+    "approval": {
+      "enabled": true,
+      "auto_approve_min": 10
+    }
+  },
+  "script": {
+    "approval": {
+      "enabled": true,
+      "auto_approve_min": 10
+    }
+  },
+  "carousel": {
+    "formats": {
+      "4:5": { "w": 1080, "h": 1350 },
+      "1:1": { "w": 1080, "h": 1080 },
+      "3:4": { "w": 1080, "h": 1440 }
+    },
+    "default_format": "4:5",
+    "slides": { "min": 2, "max": 20, "publish": 10 },
+    "min_margin_px": 64,
+    "min_font_px": 26,
+    "jpeg_quality": 2
+  },
+  "references": {
+    "dir": "../referencias"
+  },
+  "sound": {
+    "source": {
+      "name": "epidemic",
+      "search": "https://www.epidemicsound.com/sound-effects/search/?term="
+    },
+    "level_db": -10,
+    "default_voice_lufs": -29,
+    "voice_target_lufs": -29,
+    "voice_max_gain_db": 6,
+    "window_s": 0.2,
+    "max_peak_db": -1,
+    "ducking": {
+      "db": -6,
+      "before": 0.06,
+      "after": 0.25,
+      "min_gap": 0.35
+    },
+    "bed": {
+      "level_db": -24,
+      "ducking_db": -8,
+      "fade": [
+        1,
+        1.5
+      ]
+    },
+    "kit": "kit/sfx",
+    "roles": {
+      "click": {
+        "db": -4
+      },
+      "pop": {
+        "db": -4
+      },
+      "tick": {
+        "db": -8
+      },
+      "typing": {
+        "db": -8
+      },
+      "confirm": {
+        "db": -4
+      },
+      "whoosh": {
+        "db": 0,
+        "stretchable": true
+      },
+      "transition": {
+        "db": 0,
+        "stretchable": true
+      },
+      "riser": {
+        "db": 0,
+        "stretchable": true
+      },
+      "impact": {
+        "db": -2
+      },
+      "glitch": {
+        "db": -4
+      },
+      "bed": {
+        "db": 0
+      },
+      "texture": {
+        "db": 0,
+        "stretchable": true
+      }
+    },
+    "premiere_track": 1
+  },
+  "timing": {
+    "lead_in": 0.4,
+    "pause_between_lines": 0.25,
+    "final_hold": 0.8,
+    "words_per_second": 3
+  },
+  "captions": {
+    "max_chars": {
+      "9:16": 28,
+      "16:9": 42
+    },
+    "min_duration": 0.6,
+    "style": "shadow",
+    "default_position": {
+      "platform": "all",
+      "anchor": "bottom-center",
+      "x": 0,
+      "y": 0
+    },
+    "safe_area": {
+      "9:16": {
+        "shorts": {
+          "top": 0.15,
+          "bottom": 0.2,
+          "left": 0.044,
+          "right": 0.178
+        },
+        "tiktok": {
+          "top": 0.125,
+          "bottom": 0.2,
+          "left": 0.111,
+          "right": 0.111
+        },
+        "reels": {
+          "top": 0.14,
+          "bottom": 0.2,
+          "left": 0.06,
+          "right": 0.06
+        }
+      },
+      "16:9": {
+        "youtube": {
+          "top": 0.05,
+          "bottom": 0.1,
+          "left": 0.05,
+          "right": 0.05
+        }
+      }
+    }
+  },
+  "assembly": {
+    "lufs": -14,
+    "peak_db": -1,
+    "lra": 11,
+    "crf": 17,
+    "audio_kbps": 320
+  },
+  "premiere": {
+    "node": "",
+    "server": "",
+    "temp_dir": "/tmp/premiere-mcp-bridge",
+    "export_preset": "",
+    "media_encoder_installed": false,
+    "export_via_extendscript_approved": false
+  },
+  "publish": {
+    "auto": {
+      "approved": false,
+      "platforms": [
+        "youtube",
+        "instagram",
+        "tiktok"
+      ]
+    },
+    "text": {
+      "footer_youtube": "",
+      "footer_instagram": "",
+      "hashtags": []
+    },
+    "youtube": {
+      "privacy": "public",
+      "audited": false,
+      "category": "28",
+      "language": "pt-BR",
+      "made_for_kids": false,
+      "synthetic_content": false
+    },
+    "instagram": {
+      "api_version": "v23.0",
+      "share_to_feed": true,
+      "cover_at_s": null
+    },
+    "tiktok": {
+      "mode": "direct",
+      "privacy": "PUBLIC_TO_EVERYONE",
+      "audited": false,
+      "comments": true,
+      "duet": false,
+      "stitch": false,
+      "ai_content": true,
+      "cover_at_s": null
+    },
+    "schedule": {
+      "max_delay_h": 12,
+      "attempts": 3,
+      "interval_min": 10
+    }
+  },
+  "cloud": {
+    "url": "",
+    "omnichat": false,
+    "metrics": false,
+    "schedule": false,
+    "studio": false,
+    "studio_url": ""
+  },
+  "omnichat": {
+    "instagram": {
+      "interval_min": {
+        "first_48h": 1,
+        "first_week": 5,
+        "after": 30
+      },
+      "max_per_pass": 15,
+      "pause_between_s": 2,
+      "attempts": 3,
+      "rate_limit_pause_min": 30,
+      "default_dm": "Oi, @{username}! Aqui está o link que você pediu: {link}",
+      "default_public_replies": [
+        "Te mandei no direct 👀",
+        "Acabei de te mandar no direct 📩",
+        "Corre no direct! 🚀",
+        "Já está no seu direct 📬",
+        "Mandei no seu direct, dá uma olhada 👀",
+        "Olha o direct, @{username}! 📩",
+        "Te enviei no direct agora mesmo 🚀",
+        "Chegou no seu direct, confere lá 👇",
+        "Direct enviado, @{username}! 🙌",
+        "Dá uma olhada no direct, mandei por lá 📲",
+        "Acabou de cair no seu direct ✅",
+        "Mandei lá no direct pra você 🔥"
+      ],
+      "generic_phrases": [
+        "quero",
+        "eu quero",
+        "me manda",
+        "manda",
+        "me envia",
+        "envia",
+        "me passa",
+        "link"
+      ],
+      "context_model": "haiku",
+      "enable_on_publish": false,
+      "quote_post": true
+    }
+  },
+  "history": {
+    "versions": 5
+  },
+  "models": {
+    "orchestrator": {
+      "model": "sonnet",
+      "effort": "medium"
+    },
+    "reference": {
+      "model": "opus",
+      "effort": "high"
+    },
+    "stages": {
+      "av-script": {
+        "model": "opus",
+        "effort": "high"
+      },
+      "av-motion": {
+        "model": "opus",
+        "effort": "high"
+      },
+      "av-narration": {
+        "model": "haiku",
+        "effort": "high"
+      },
+      "av-voice": {
+        "model": "haiku",
+        "effort": "high"
+      },
+      "av-render": {
+        "model": "haiku",
+        "effort": "medium"
+      },
+      "av-captions": {
+        "model": "haiku",
+        "effort": "low"
+      },
+      "av-assemble": {
+        "model": "haiku",
+        "effort": "medium"
+      },
+      "av-premiere": {
+        "model": "haiku",
+        "effort": "medium"
+      },
+      "av-thumbnail": {
+        "model": "sonnet",
+        "effort": "xhigh"
+      },
+      "av-analysis": {
+        "model": "sonnet",
+        "effort": "high"
+      },
+      "av-carousel": {
+        "model": "opus",
+        "effort": "high"
+      }
+    }
+  },
+  "agents": {
+    "accounts": [{ "id": "claude-personal", "provider": "claude", "dir": "~/.claude-personal" }],
+    "codex": { "model": "gpt-6-luna", "effort": "medium", "sandbox": "auto" }
+  },
+  "studio": {
+    "claude": {
+      "bin": "claude",
+      "permission_mode": "auto",
+      "budget_usd": null,
+      "bash_timeout_ms": 7200000,
+      "allow": [
+        "Read",
+        "Agent",
+        "Edit",
+        "Write",
+        "Glob",
+        "Grep",
+        "Skill",
+        "WebFetch",
+        "WebSearch",
+        "TodoWrite",
+        "Bash(node tools/*)",
+        "Bash({python} tools/*)",
+        "Bash(ffprobe *)",
+        "Bash(ffmpeg *)",
+        "Bash(mkdir *)",
+        "Bash(cp *)",
+        "mcp__claude-in-chrome"
+      ]
+    }
+  },
+  "agent": {
+    "engine": "grok",
+    "wait_max_min": 20
+  },
+  "admin": {
+    "emails": []
+  }
+}

package/package.json

@@ -1,6 +1,42 @@
 {
   "name": "@crievo/omni",
-  "version": "0.0.0-stage",
-  "stub": true,
-  "description": "Temporary package placeholder for staged publishing"
-}
+  "version": "0.1.1",
+  "description": "OmniVideo no seu computador: o CLI omni, o studio local e as ferramentas do pipeline.",
+  "type": "module",
+  "bin": {
+    "omni": "bin/omni.mjs"
+  },
+  "engines": {
+    "node": ">=22"
+  },
+  "os": [
+    "darwin"
+  ],
+  "license": "UNLICENSED",
+  "homepage": "https://omni.crievo.com",
+  "dependencies": {
+    "@hono/zod-validator": "0.9.1",
+    "@hugeicons/core-free-icons": "^4.3.5",
+    "@hugeicons/react": "^1.1.10",
+    "@radix-ui/react-accordion": "^1.3.1",
+    "@radix-ui/react-checkbox": "^1.4.1",
+    "@radix-ui/react-popover": "^1.3.1",
+    "@radix-ui/react-select": "^2.4.1",
+    "@radix-ui/react-slider": "^1.6.1",
+    "@radix-ui/react-switch": "^1.4.1",
+    "@radix-ui/react-tooltip": "^1.4.1",
+    "@tanstack/react-query": "^5.104.3",
+    "@tanstack/react-router": "^1.170.44",
+    "border-beam": "1.4.1",
+    "hono": "4.13.9",
+    "marked": "^18.1.0",
+    "react": "19",
+    "react-day-picker": "^10.0.2",
+    "react-dom": "19",
+    "react-is": "19",
+    "recharts": "3.10.1",
+    "voice-glow": "0.3.0",
+    "zod": "4.6.5",
+    "bun": "1.4.3"
+  }
+}

package/reference/captions.md

@@ -0,0 +1,156 @@
+# Captions
+
+`omnivideo.py captions` applies these rules. Change them here and in the code together.
+
+## Text
+
+- **Spelling comes from the written line.** Nothing is transcribed: the time of each word is measured on the audio.
+- **A trailing period, comma, semicolon or colon is dropped from a caption**; question marks,
+  exclamation marks and ellipses stay. A comma in the middle stays.
+- **One line per caption.** At most 28 characters in 9:16 and 42 in 16:9
+  (`captions.max_chars` in the config).
+- **A caption's text is edited in the studio**, in the list of the studio's
+  Legenda (Captions) stage (double-click or the pencil; Enter saves, Esc
+  cancels). The timing does not change: only the words. The new text goes
+  straight into `captions.srt` (the previous one stays in the stage history)
+  and `captions.texts` in `scenes.json` records the swap,
+  `{ "<line>.<n>": { "text", "original" } }`, where `original` is what the
+  line generated. Generating the captions again restores the edited text to
+  the caption of a line that still generates the same `original`; if the line
+  changed, the edit is dropped and `omnivideo.py captions` says which one.
+  Going over the character limit does not block: the list shows the count in
+  orange and the template wraps to two lines. An edited caption is marked
+  "Editada" (Edited), and the field's ↺ goes back to the line's text.
+
+## Breaking
+
+1. **Sentence first.** Break at `.`, `?`, `!`, `…`, `;`, `:` and comma.
+2. **Join a short comma piece** with the next one if it fits: "O Jev, zero".
+3. **A long sentence is split into balanced parts**, with a cost for:
+   - a caption over the limit (forbidden);
+   - ending on a linking word: de, da, do, a, o, e, em, um, que, com, pra,
+     no, na, por, se, seu, ao, à, é, ou, mas;
+   - ending on a number (it would separate "70" from "a 500");
+   - a single word alone in a caption.
+4. **Never spans two lines (spoken lines).** Captions follow each line's timing.
+
+Examples that came out like this and were fixed:
+
+| bad | good |
+|---|---|
+| "Cada decisão leva de 70" / "a 500 milissegundos" | "Cada decisão leva" / "de 70 a 500 milissegundos" |
+| "Uma IA de ponta pode levar" / "minutos pra tomar uma única" / "decisão" | "Uma IA de ponta pode" / "levar minutos pra tomar" / "uma única decisão" |
+| "quase duas ordens de" / "grandeza mais barato" | "quase duas ordens" / "de grandeza mais barato" |
+
+## Timing
+
+- Starts at the start of the first word and ends at the end of the last one.
+- If the next caption starts less than 0.35 s later, this one stretches up to
+  it (avoids flicker).
+- Minimum duration 0.6 s, without running into the next one.
+
+## Style
+
+Captions are burned by `tools/assemble.mjs`: each cue of `captions.srt` is
+drawn in `templates/captions.html`, becomes a transparent PNG and is overlaid
+from its start to its end. The ffmpeg on this machine has neither libass nor
+drawtext, and doesn't need them: the style is CSS, like the scenes'.
+
+- `shadow` (default, from `captions.style` in the config): Schibsted Grotesk
+  Bold, white with a soft shadow, for dark backgrounds.
+- `pill`: dark text on a translucent white pill with a light shadow. Reads on
+  both light and dark backgrounds.
+- 47 px type at 1080 width: 28 characters fit in the narrowest safe area
+  (`all`). If it doesn't fit, it wraps into two balanced lines.
+- **Color, outline and shadow** are keys of `captions` in `scenes.json`, at
+  the same three levels as the position (video → scene → caption). A key that
+  is in no level keeps the style default:
+
+  | key | what it is | range |
+  |---|---|---|
+  | `color` | text color (`#rrggbb`) | pill `#111111`, shadow `#ffffff` |
+  | `background`, `background_opacity` | pill color and opacity (only in the `pill` style) | `#ffffff`, 0 to 1 (0.9) |
+  | `outline`, `outline_color` | text outline width, in frame px, and color | 0 to 40 (0 = none) |
+  | `shadow_color`, `shadow_opacity` | shadow color and opacity | `#000000`, 0 to 1 |
+  | `shadow_blur`, `shadow_x`, `shadow_y` | blur and offset, in frame px | 0 to 100; −100 to 100 |
+
+  The shadow belongs to the pill in the `pill` style and to the text in
+  `shadow`. Touching any `shadow_*` key replaces the style's shadow (two
+  layers in `shadow`) with a single layer built from the five values. The
+  outline is painted behind the text, so the width is only the outer part.
+  The px defaults (`styleDefaults`, in `tools/captions.mjs`) scale with the
+  frame width.
+- A tweak that applies to several videos or that the panel doesn't cover
+  (type size, weight, font) goes in `motion/src/_captions.css`, appended to the
+  end of the layer's style. Color, outline and shadow set in `scenes.json` win
+  over it (inline style).
+
+## Position
+
+In `captions` in `scenes.json`, at three levels, each one over the previous:
+
+```json
+"captions": {
+  "platform": "all",
+  "anchor": "bottom-center",
+  "x": 0, "y": 0,
+  "style": "shadow",
+  "safe": { "bottom": 0.12 },
+  "color": "#ffd400", "outline": 4, "shadow_opacity": 0.5,
+  "scenes":     { "c04": { "anchor": "top-center", "y": 40 } },
+  "individual": { "c03f1.2": { "x": -30 } }
+}
+```
+
+- **Safe area** (`platform`): the area no interface covers. It is the
+  reference for the anchor and for the warning, not a lock: the caption goes
+  wherever x/y send it, including outside it, and the studio warns in orange
+  on which side and by how many px it went over. Only the frame limits it.
+  Values in `captions.safe_area` in the config, as a fraction of the frame:
+
+  | 9:16 | top | bottom | left | right |
+  |---|---|---|---|---|
+  | `shorts` | 15 % | 20 % | 4.4 % | 17.8 % |
+  | `tiktok` | 12.5 % | 20 % | 11.1 % | 11.1 % |
+  | `reels` | 14 % | 20 % | 6 % | 6 % |
+  | `all` | the largest on each side: 15 % · 20 % · 11.1 % · 17.8 % | | | |
+
+  The bottom used to be 35 % and ate the frame; 20 % puts the caption's bottom
+  edge at y ≈ 1536 and is the same band `templates/scene.html` reserves for it.
+- **Safe area adjustment** (`safe`): per video, only the sides you list
+  (`top`, `bottom`, `left`, `right`, a fraction of the frame, 0 to 0.8)
+  replace the platform's value. The studio edits this under Margens (Margins),
+  in Todas (All); switching the platform resets the adjustment.
+
+  In 16:9, `youtube`: 5 % at the top and sides, 10 % at the bottom (player bar).
+  Sources: 2026 safe zone guides (reap.video, upload-post, adaptlypost);
+  check them now and then, the interfaces change.
+
+- **Anchor** (`anchor`): 3×3 matrix, `top|middle|bottom` - `left|center|right`.
+  Pins the caption box to the edge (or the center) of the safe area:
+  `bottom-center` puts the caption's bottom edge on the safe area's bottom edge.
+- **Offset** (`x`, `y`): frame px from the anchor. Positive `x` goes right,
+  positive `y` goes down; negative, the opposite.
+- **Levels:** the video level applies to all captions; `scenes.<id>` changes
+  only that scene's captions (a scene with the hero at the bottom, for
+  example); `individual.<id>` changes a single caption. The id is
+  `<line>.<n>`, the line's n-th caption.
+- **Turning a caption off** (`individual.<id>.hidden: true`): the caption
+  stays in `captions.srt` and in the list, but `assemble.mjs` doesn't burn it
+  (it reports how many were left out). The studio toggles it with the "Mostrar
+  esta legenda no vídeo" (Show this caption in the video) switch, which
+  appears when a caption is selected (it applies to all the selected ones);
+  the preview shows a hidden caption dimmed. On is the default: the key only
+  exists when off. Only at the caption level, not the scene level.
+
+In the `all` safe area, `bottom-center` puts the caption's bottom edge at
+y ≈ 1536 (20 % above the bottom of the frame), next to the band
+`templates/scene.html` reserves (`--caption-zone`, 19 %). Videos made when the
+bottom was 35 % had the caption at y ≈ 1248; when the assembly is redone it
+moves up less, and a negative `captions.y` that was meant to compensate now
+pushes the caption too high: review it.
+
+Where to edit: the studio's Legenda (Captions) stage (preview over the real
+frame, matrix, X/Y, drag) or `scenes.json` directly. `node tools/assemble.mjs
+videos/<slug> --sheet` shows the captions over one frame of each scene without
+assembling the video.