Magnific workflow adds audio-reference lip sync for Seedance

• techhalla's Seedance lip sync workflow post and prompt follow-up show a Magnific setup built around feeding audio into Seedance as a reference, instead of keyframing mouth timing clip by clip techhalla's workflow post and techhalla's prompt follow-up.
• The useful part is distribution, not just the demo: techhalla's shared Space opens the whole canvas as a duplicable Magnific Space, with prompts and assets intact.
• DavidmComfort's test clip adds a second example from outside the original thread, claiming Seedance can get strong lip sync from an audio file alone, while DavidmComfort's setup post says the setup is simple enough to show in one screenshot.
• Magnific's own examples are already leaning on the same stack: Magnific's taste thread said the stills came from GPT2 and the motion came from Seedance 2.0, and techhalla's earlier Space post said that workflow work happens before generation, with 25-plus prompts packed into a reusable Space.

You can open Magnific, jump straight into techhalla's shared lipsync Space, and compare it with a separate audio-file test from DavidmComfort. Magnific's own earlier showcase framed the appeal as taste over speed, while techhalla's older Space breakdown exposed the production detail most people skip: stills, animation model, prompt pack, and the choice to stay at 720p to keep costs down.

The core claim in techhalla's workflow post is straightforward: Seedance lip sync gets more reliable when the workflow passes in audio as reference material. In the follow-up, techhalla says the shared file includes the exact prompts and assets used for the demo, and specifically calls out instructions telling Seedance to use the audio tracks as reference.

That changes the shape of the task. The interesting bit is less about one polished clip, and more about turning lip sync into a promptable input condition inside a repeatable canvas.

The delivery format matters here. Rather than posting a recipe in screenshots, techhalla's shared Space post invites people to duplicate the entire Magnific Space, remix the prompts, and reuse the assets.

For creative tooling, that is a better artifact than a thread. A Space preserves the stack in one place: source material, prompt wording, and the order of operations.

A separate post from DavidmComfort makes the same point with less packaging. He wrote that Seedance can do "really good Lip Sync" from an audio file, then followed with a setup post showing the configuration behind the result.

That second example matters because it narrows the claim. The technique is not only "this one Magnific template worked." It also appears to work as a lighter setup built around an audio file itself.

The broader workflow picture showed up before the lip sync thread landed. In techhalla's earlier post, he said more than 80 percent of his time goes into refining the workflow before generating anything, and the thread context lists the stack plainly:

• Nano Banan Pro and 2 for still images, per techhalla's earlier Space post
• Seedance 2.0 for animation, according to the same thread
• 720p output to keep costs down, also from that workflow note
• A Magnific Space containing 25-plus prompts, again in techhalla's post

Magnific's own showcase post matches that split. Magnific's thread says the images came from GPT2 and the videos from Seedance 2.0, which makes the lip sync posts feel like a specific extension of an already standard stack: generate the still, animate in Seedance, then tighten speech timing by handing the model audio reference data.