Update README.md

README.md

@@ -20,13 +20,13 @@ tags:
 > [Wan2.1](https://github.com/Wan-Video/Wan2.1) release.
 
 A 1.3 B student distilled from a Wan2.1-14B teacher that generates video
-**continuously** rather than as a fixed clip: 750 ms of 640×368 video per block,
-generated in 428 ms on one A100-80GB, extended block by block for as long as you
+**continuously** rather than as a fixed clip: 750 ms of 640x368 video per block,
+generated in 428 ms on one A100 80GB, extended block by block for as long as you
 let it run. The text conditioning can be changed mid-stream without clearing the
 K/V cache, so the scene continues instead of cutting.
 
 **Code, setup and documentation: [JonathanColetti/LiveWan](https://github.com/JonathanColetti/LiveWan).** This repository holds
-weights and data only; it is not usable on its own.
+weights and data only. It is not usable on its own.
 
 <p align="center"><img src="samples/world_p60_waterfall_16fps.mp4" width="480"></p>
 
@@ -58,7 +58,7 @@ download these unless you intend to train.
 | `checkpoints/t14b_b64/opt_rank00{0..7}.pt` | 8 × 2.2 GB | FSDP-sharded AdamW state at step 3000, one shard per rank |
 
 Nothing here duplicates a public download. The Wan2.1 base model, its VAE and the
-umt5-xxl encoder are **not** mirrored in this repository — `setup.sh` pulls
+umt5-xxl encoder are not mirrored in this repository. `setup.sh` pulls
 [Wan-AI/Wan2.1-T2V-1.3B](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B) (17 GB)
 from its own repo, and the 14B teacher is only needed if you are distilling.
 
@@ -156,13 +156,7 @@ shows up as a shifted mean or changed variance; neither appeared.
 - **The batch-64 premise is unsettled** — confounded with having simply trained
   longer.
 
-## Do not re-encode the prompts
 
-`data/prompts.pt` is the conditioning every clip in this project was generated and
-trained under. umt5-xxl embeddings are **hardware-dependent**: the same strings
-encoded on an A100 versus an H200 differ by 2.2% relative, while two runs on the
-same box agree to 4.9e-09. Re-encoding silently decouples the conditioning from the
-results and nothing will look wrong.
 
 ## Licence