arkennemasis Avatar Brief (the idea form)
The idea form that starts a fully-automated presenter video
- brief
- source_music
- source_entertainment
- override_url
- seconds
- words
ArkAvatarBrief is the front door to the arkennemasis Avatar system - the six nodes that turn "cover today's news about X" into a captioned vertical video with nobody watching. Where the other avatar nodes read and write, this one is the idea form. You type what you actually want, and it assembles the prompt for the script-writing model, hands the story hunt its source pages, and works out how many words your target length allows.
It's a pure prompt-builder node - no model, no API call, nothing paid. You fill in a box per decision, it emits the assembled brief string plus a few ready-to-wire values. That's the whole job, and it's the most important node in the pipeline for a boring reason: everything downstream is only as good as what you type here.
How it works
The required inputs are the things you genuinely decide. source_music and source_entertainment are the source pages the story hunt screenshots - music first, because the pack's own framing is that a strong music story beats an entertainment one, and source_entertainment is only read when the music candidates are thin. Sources go one per line, and prefixing a line with # switches that source off without losing it - handy when one publisher is being blocked or is down today.
The inputs that matter most for a beginner:
topic- what this channel covers and who watches it. It steers which part of a story is worth 30 seconds.presenter- who is speaking and how: their register, pace, attitude. Read the tooltip carefully: this is the voice of the script, not the face. The face comes from the reference photo elsewhere in the graph.seconds- the target spoken length. The word budget derives from it, and so does render cost, since video time scales with frames.language- 11 choices, from English and Hindi through Japanese and Arabic. It sets both the script and the caption.
The optional inputs are where you add seasoning. call_to_action closes the video (blank gets the model's usual generic "follow for more"), must_mention forces a name or number in, avoid keeps words and claims out, and special_request is emitted first in the prompt because a model weights what it reads first. override_url is the escape hatch: make today's video about a specific article and skip the automatic pick entirely - the day the model chooses badly.
Outputs
The brief output is the assembled prompt, wired into the script-writing LLM. source_music and source_entertainment go to the story-hunt side, override_url passes through so the picker knows you've overridden, and seconds plus the derived words give downstream nodes the number to plan around.
Installing
Like everything else in this pack:
cd ComfyUI/custom_nodes
git clone https://github.com/Hishamahmer/comfyui-arkennemasis
pip install replicate httpx
Then restart ComfyUI. Or find arkennemasis in ComfyUI Manager and let it handle both. This node itself needs no extra models or keys - the pack's other paths do, but modules load independently, so this one just works.
Troubleshooting
The classic failure is upstream of the node, not inside it: a blank topic or presenter produces a perfectly grammatical, completely generic video, because you asked for one. The pack's philosophy is "you bring the words, the pack brings the machine" - nothing here tells a model what to write. And if the automatic story pick keeps missing, that's what override_url is for. The pack bills you for renders, not for this node, so re-running a brief is free.
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| source_music | STRING | https://www.musicbusinessworldwide.com/ https://completemusicupdate.com/ https://www.musicweek.com/ | MUSIC sources, ONE PER LINE. The primary beat — a strong music story always beats an entertainment one. Prefix a line with # to switch a source off without losing it. Every source here was verified by screenshotting a real article from it. |
| source_entertainment | STRING | https://variety.com/v/music/ | The ENTERTAINMENT section page, read second and used only when the music candidates are thin. |
| topic | STRING | What this channel covers and who watches it. Steers which part of the story is worth 30 seconds. | |
| presenter | STRING | Who is speaking and how — their register, pace and attitude. This is the voice of the script, not the face; the face comes from the reference photo. | |
| seconds | FLOAT | 305–180 | Target spoken length. The word budget is derived from this, and so is the cost of the render — video time scales with frames. |
| language | COMBO | The language of the script and the caption. | |
| call_to_actionopt | STRING | How it should close. Left blank the model writes its own, which is usually a generic 'follow for more'. | |
| must_mentionopt | STRING | Anything that has to be in there — a name, a number, a product. | |
| avoidopt | STRING | Words, claims or angles to keep out. | |
| special_requestopt | STRING | The one thing that matters most today. Emitted FIRST, because a model weights what it reads first. | |
| override_urlopt | STRING | Make today's video about THIS article instead, skipping the automatic pick. For the day the model chooses badly, or when you already know the story you want. |
Outputs (6)
| Name | Type | Description |
|---|---|---|
| brief | STRING | — |
| source_music | STRING | — |
| source_entertainment | STRING | — |
| override_url | STRING | — |
| seconds | FLOAT | — |
| words | INT | — |