Sign In

Minimax H3 SEED HUNTER - Latent Upscaler + Seamless Video Continuation + Speedups + Ref Mods - v2.5 Showcase

Loading Images

Are you sick of your dialogue being butchered by h3? Well here you go: Seed Hunting for dialogue! Easy toggle on/off option, no new models necessary. All of my new videos use it -- it takes the guesswork out of audio so you can focus all your genning on the visual!

v2.5

  • Adds the Dialogue / Voice Clone Tool (quickly generate 3x line reads for use in your gens -- super user friendly and easy to toggle on and off without ever unloading h3)

  • Redesigns the workflow (many people said the workflow felt too claustrophobic, so I reorganized and spread things out a bit)

  • Small additions: 1. Adds clickable links instead of notes for all model / node / lora resources, 2. Block/Sparse attention can now be toggled on and off for both 1st pass AND 2nd pass individually, so Stage 1 gens are now faster by default.

To try the new Dialogue / Voice Clone Tool: just toggle it on in the Gen Options, and literally click run!~ Without any voices loaded, it'll create an AI generated voice based on what's written in the "Vocal Tone & Delivery" text box. But I recommend you try loading a voice file, you can do so either in <Audio 1> or in the Custom Audio loader inside the group. Gen your character's line of dialogue, toggle the tool off, throw the new line of dialogue into your ref loader's audio slot, and now you can use that line of dialogue in your next gen without having to worry about the character babbling, misspeaking, etc.

Comments