Skywork AI Skyreels V4 Text To Video (Official Edition)

Every model family has its entry-point node, and for Skyreels V4 this is it: pure text-to-video, no reference images, no anchor frames, no @-tag syntax. If you've never touched the Skyreels nodes in BizyAirPlus, start here before you go juggling reference conditioning.

What this node does

BizyTRD_Skyreels_V4_Official_T2V_API turns a prompt into a 3–15 second clip at up to 1080p, with synchronized audio on by default. It's the baseline of the V4 family - the R2V and FLF nodes add character/reference control on top, but this one is just "describe it and go."

Like every node in this pack it's cloud execution. Flip the BizyAirPlus button ON, and Skyreels V4 runs on BizyAir's servers; ComfyUI uploads the prompt, streams back progress, and drops the finished clip into your graph. No local model, no VRAM requirements, no checkpoint hunt.

The inputs that matter

• prompt - your whole job. V4 is a modern video model, so natural language with concrete camera language works well: "slow push-in on a rainy street, neon reflections, a figure with an umbrella walks into frame."
• resolution - 1080p by default. That's the ceiling in the combo.
• duration - 3 to 15 seconds.
• mode - std (default) or fast. Fast is snappier but currently only works with sound off; it doesn't support audio-video output yet.
• sound - on by default, which is the differentiator: V4 generates dialogue and sound effects, not silent footage. Flip off for a mute take.
• aspect_ratio - 16:9 by default.

The two shared BizyAir options are there too: api_config (override base URL/key for one request) and skip_error (emit an error placeholder instead of killing the whole workflow when a cloud call fails). Leave them alone until you need them.

Outputs

• video - the finished clip, wired into a preview or save node.
• request_id - the cloud task ID for logs.
• urls - the hosted result URLs, if you want the file directly.

How to install

cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt

Restart ComfyUI (or install "BizyAirPlus" via ComfyUI Manager). Dependencies are bizyair-cloudberry and bizytrd; nothing heavy downloads because the model runs remotely. First run: account at bizyair.ai → API key → enter it when the button prompts (Settings > BizyAirPlus > API Key also works; stored at ~/.BizyAirPlus/apikey.ini).

Where people get burned

The fast mode / sound conflict is the recurring one - leave sound on and flip to fast, and you get a failure or silence. If you want both speed and audio, you can't have it yet on V4.

The other thing is expectations: this is T2V, so character consistency across the clip is whatever the model decides. If you need a specific face or object to hold, that's what the V4 R2V node (with its @image1 reference tags) is for. And on the pack itself: BizyAir's malware reputation is leftover from a 2025 ComfyUI Manager auto-install bug, not reality - it's a legitimate SiliconFlow pack, and if you ever remove it, use pip uninstall bizyair-cloudberry bizytrd, not folder deletion alone.