Wan Bailian CN I2V Config

Wan the video model has a split personality, and this node lives on the newer side of it. The Wan 2.1/2.2 you see everywhere in local ComfyUI workflows is open-weights; the Wan 2.7 this node calls is API-only, served by Alibaba's Bailian/DashScope cloud. WanBailianCNI2VConfig turns a prompt plus a first frame into a video request for wan2.7-i2v, which is the current closed I2V model - a model with no local equivalent you can download.

As with every video config here, it outputs a MEDIAHUB_VIDEO_REQUEST, not a video. The chain: config → MediaHubSubmitVideo → MediaHubResolveVideo → SaveVideo, with a DashScope key supplying the authentication.

Inputs that matter

• mode - first_frame or first_last_frame. Same binary choice as the Seedance I2V node: animate from a still, or bookend a clip between two images.
• first_frame_image - required; JPEG-encoded and uploaded with the request.
• last_frame_image - optional, only used in first_last_frame mode.
• model - a single choice, wan2.7-i2v-2026-04-25 (the date stamp is a release marker, not a version to fiddle with).
• resolution - 720P or 1080P.
• duration - default 5 seconds, range 2–15.
• prompt_extend - on by default. Alibaba's prompt-enhancement pass runs server-side before generation. Like MiniMax's video optimizer, it usually helps; turn it off if you need your exact words to reach the model.
• watermark - off by default.
• seed - -1 = random.
• audio_url (optional) - a URL to an audio track you want the video to match or carry. This is a distinctive Wan 2.7 feature; drop in a music or narration URL and the model works with it rather than you splicing audio later.

Output: request → MediaHubSubmitVideo.

Why you'd reach for it

Because it's the newest Wan video, and Wan's reputation for solid, controllable motion is exactly why people kept coming back to its local models. If you want that lineage's current iteration and can't (or won't) run it locally - because you can't - this is the door. The audio_url input is a genuine differentiator nobody else in the pack matches.

Installing it

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

Restart ComfyUI; under MediaHub/Provider/Wan Bailian CN/Video. Dependencies requests/Pillow/numpy, plus a DashScope API key (same key as the Wan/HappyHorse image nodes).

Common gotchas

Don't confuse this with local Wan - different model, different billing, different everything. And keep the source frame's aspect ratio aligned with your resolution choice; Wan is less forgiving than some providers about mismatch. audio_url expects a URL, not a local file path - if you have a file, host it or skip the input and splice audio yourself. And the usual async rules: Submit alone gives you a job, not a clip, and ResolveVideo can legitimately poll for minutes.