ArchAi3D_Qwen_Person_Cinematographer

ArchAi3D_Qwen_Person_Cinematographer is the pack's "direct the shot" node for people, and it has the most intimidating input list in the whole set - emotion, eyes, pose, lighting, mood, orbit angle, distance - until you notice the default preset is a thing called ascendra_man_threatening and realize this thing comes with cinematic scenarios pre-built. The ASCENDRA presets are named little film moments; pick one and a wall of dropdowns configures itself.

Which is exactly the pitch: a cinematographer in a box for Qwen-Edit. It moves the camera around a character (orbit is the headline movement, with orbit_angle up to 360°) while the generated system prompt works to keep the character's face, clothes, hairstyle and core pose intact - the hard problem for any edit model, and the one Qwen-Edit is honestly weakest at without help.

The inputs that matter

The full list is genuinely long (emotion with 20 options, eyes with 17, pose with 25+, 8 lighting setups, 9 moods). You will not touch most of it. What matters:

• preset - the cinematic scenarios, ascendra_man_threatening by default. This is the "why did I install this" feature: one click, done.
• scene_context - the character and setting ("hooded man in dark jacket on train platform at night"). This is the input that carries most of the quality.
• movement + orbit_angle - camera path and how far it travels. Orbit is the most reliable movement for Qwen, per the pack's research.
• multi_frame_mode + steps - for 360° turnarounds or emotion cycles. Same chain-edits pattern as the rotation nodes: each frame is a full Qwen-Edit pass.

Outputs: prompt and system_prompt (the two you wire into the encoder), plus multi_frame_prompts, frame_count, preset_description, and a technical_notes string that gives tips for the selected shot - a nice touch you don't get on the other camera nodes.

How it works

Underneath, it assembles a director-style instruction - movement, angle, emotion, eye direction, pose, lighting, mood - into one sentence for the edit model, wrapped with a system prompt that tells the model it's a professional cinematographer preserving the actor's identity across the camera move. Identity preservation is the load-bearing part: Qwen-Edit re-emits the whole frame, so without explicit "keep the face identical" language, your character becomes someone else between angles.

Installing it

Inside ComfyUI-ArchAi3d-Qwen by Amir Ferdos (ArchAi3d):

# ComfyUI Manager: search "ArchAi3d Qwen"
cd ComfyUI/custom_nodes
git clone https://github.com/amir84ferdos/ComfyUI-ArchAi3d-Qwen.git
cd ComfyUI-ArchAi3d-Qwen
pip install -r requirements.txt

Restart. You supply the Qwen-Image-Edit model (GGUF-quantized for consumer VRAM).

Common issues

• Faces drift. The system prompt helps, but the KB and the community agree: Qwen-Edit struggles most with faces it hasn't seen in training. For a character that must stay identical, expect to inpaint the face back afterward. Strict identity work is still "inpaint a face back in" territory.
• The model may move the person, not the camera. When people are in frame, Qwen sometimes re-poses them instead of orbiting the camera. Centered subject, short orbit angles, and the system prompt are your levers.
• Multi-frame chains drift. Each frame builds on the last; errors compound. Regenerate from the source image rather than the previous output if a sequence wanders.

It's the most "fun" node in the pack - the preset list alone invites experimentation - and it's the right tool for character-focused camera moves, with the honest caveat that identity preservation is prompt-assisted, not guaranteed.