Ovi is Local Version of VEO 3 & SORA 2 - The first-ever public, open-source model that generates both VIDEO and synchronized AUDIO, and you can run it on your own computer on Windows even with a 6GB GPUs - Full Tutorial for Windows, RunPod and Massed Compute - Gradio App

will it be updated to ovi 1.1 ( trained to up to 10s videos, 960x960 and 100% more videos !). there is a Prompt Format Update too for this version so each version need is own window I think.

Amazing. Its so much faster than the standard GUI.

sir it has stopped working on Massed Compute

sir OVI has stopped working. OVI_PRo_8 shows error from past 1 week

Hello my friend. First of all, thank you for providing such great information! I ran into an error upon running my first example video. Can you look at the text and determine what I may have missed. I;m new to this, so I appreciate any help. [STARTUP] Set MKLOMP threads to 12 for optimal CPU performance [DEBUG] Parsed args single_generation=False, single_generation_file=False, test=False, test_subprocess=False Starting Gradio interface with lazy loading... Share mode DISABLED (local only) Use --share flag to enable public access with a shareable URL Ovi Pro SECourses Premium App v8.3 [DEBUG] Main block single_generation_file=None, single_generation=None, encode_t5_only=False, test=False, test_subprocess=False Running on local URL http127.0.0.17860 To create a public link, set `share=True` in `launch()`. Auto-loading last used preset 24-GB GPUs - Faster FP8 Scaled ============================================================ GPU DETECTION RESULTS GPU Model NVIDIA GeForce RTX 4090 GPU Count 1 VRAM Size 23.99 GB ============================================================ ============================================================ SYSTEM RAM DETECTION RESULTS Total RAM 15.94 GB ============================================================ ✓ RAM optimization Enabled Clear All Memory (RAM 15.9GB 128GB) [PRESET] Applied automatic optimizations for '24-GB GPUs - Faster FP8 Scaled' RAM 15.9GB 128GB → Enabled Clear All Memory ============================================================ GPU DETECTION RESULTS GPU Model NVIDIA GeForce RTX 4090 GPU Count 1 VRAM Size 23.99 GB ============================================================ ============================================================ SYSTEM RAM DETECTION RESULTS Total RAM 15.94 GB ============================================================ ✓ RAM optimization Enabled Clear All Memory (RAM 15.9GB 128GB) [PRESET] Applied automatic optimizations for '24-GB GPUs - Faster FP8 Scaled' RAM 15.9GB 128GB → Enabled Clear All Memory ============================================================ GPU DETECTION RESULTS GPU Model NVIDIA GeForce RTX 4090 GPU Count 1 VRAM Size 23.99 GB ============================================================ ============================================================ SYSTEM RAM DETECTION RESULTS Total RAM 15.94 GB ============================================================ ✓ RAM optimization Enabled Clear All Memory (RAM 15.9GB 128GB) [PRESET] Applied automatic optimizations for '24-GB GPUs - Faster FP8 Scaled' RAM 15.9GB 128GB → Enabled Clear All Memory [LORA DEBUG] Received from UI lora_1='None', lora_1_scale=1, lora_1_layers='Video Layers' lora_2='None', lora_2_scale=1, lora_2_layers='Video Layers' lora_3='None', lora_3_scale=1, lora_3_layers='Video Layers' lora_4='None', lora_4_scale=1, lora_4_layers='Video Layers' lora_specs=None [LORA STATUS] UI selections will be processed 4 LoRA(s) [1] None (scale 1) [2] None (scale 1) [3] None (scale 1) [4] None (scale 1) [RESOLUTION] Using exact user-specified resolution 960x544 ================================================================================ VIDEO GENERATION STARTED enable_multiline_prompts False enable_video_extension False Text prompt A man in a workshop, with soldering irons and circ... Image path None Resolution 544x960 Base Resolution 720x720 Duration 5 seconds Seed 99 Num generations per prompt 1 Video extensions 0 Valid prompt lines detected 1 ================================================================================ ================================================================================ CLEAR ALL MEMORY ENABLED Main process will NOT load any models All generations will run in separate subprocesses VRAMRAM will be completely cleared between generations ================================================================================ [PROMPT 11] Processing A man in a workshop, with soldering irons and circ... [GENERATION 11] Starting with seed 99 [GENERATION LORA] No LoRAs applied (using base model only) [SUBPROCESS DEBUG] Passing lora_specs to subprocess [] [SUBPROCESS DEBUG] lora_specs type class 'list', length 0 [SUBPROCESS] Running generation in subprocess... [SUBPROCESS] Command COvi_Pro_v8Ovi_ProvenvScriptspython.exe COvi_Pro_v8Ovi_Propremium.py --single-generation-file COvi_Pro_v8Ovi_Protmpr4tva3ao.json [SUBPROCESS] Params file COvi_Pro_v8Ovi_Protmpr4tva3ao.json [DEBUG] Parsed args single_generation=False, single_generation_file=True, test=False, test_subprocess=False Starting Gradio interface with lazy loading... Share mode DISABLED (local only) Use --share flag to enable public access with a shareable URL Ovi Pro SECourses Premium App v8.3 [DEBUG] Main block single_generation_file=COvi_Pro_v8Ovi_Protmpr4tva3ao.json, single_generation=None, encode_t5_only=False, test=False, test_subprocess=False [DEBUG] Taking single_generation_file path [SINGLE-GEN] Loaded params from file COvi_Pro_v8Ovi_Protmpr4tva3ao.json [SINGLE-GEN] Starting generation with params ['text_prompt', 'image', 'video_frame_height', 'video_frame_width', 'video_seed', 'solver_name', 'sample_steps', 'shift', 'video_guidance_scale', 'audio_guidance_scale', 'slg_layer', 'blocks_to_swap', 'video_negative_prompt', 'audio_negative_prompt', 'use_image_gen', 'cpu_offload', 'delete_text_encoder', 'fp8_t5', 'cpu_only_t5', 'fp8_base_model', 'use_sage_attention', 'no_audio', 'no_block_prep', 'num_generations', 'randomize_seed', 'save_metadata', 'aspect_ratio', 'clear_all', 'vae_tiled_decode', 'vae_tile_size', 'vae_tile_overlap', 'base_resolution_width', 'base_resolution_height', 'duration_seconds', 'auto_crop_image', 'base_filename', 'output_dir', 'text_embeddings_cache', 'enable_multiline_prompts', 'enable_video_extension', 'disable_auto_prompt_validation', 'force_exact_resolution', 'lora_specs', 'merge_loras_on_gpu'] [SINGLE-GEN] Text prompt A man in a workshop, with soldering irons and circ... [LORA DEBUG] Received from UI lora_1=None, lora_1_scale=1.0, lora_1_layers='Video Layers' lora_2=None, lora_2_scale=1.0, lora_2_layers='Video Layers' lora_3=None, lora_3_scale=1.0, lora_3_layers='Video Layers' lora_4=None, lora_4_scale=1.0, lora_4_layers='Video Layers' lora_specs=[] [LORA STATUS] No LoRAs selected [LORA] Using pre-built lora_specs from subprocess 0 LoRA(s) [RESOLUTION] Using exact user-specified resolution 960x544 ================================================================================ VIDEO GENERATION STARTED enable_multiline_prompts False enable_video_extension False Text prompt A man in a workshop, with soldering irons and circ... Image path None Resolution 544x960 Base Resolution 720x720 Duration 5 seconds Seed 99 Num generations per prompt 1 Video extensions 0 Valid prompt lines detected 1 ================================================================================ ================================================================================ INITIALIZING OVI FUSION ENGINE IN MAIN PROCESS Block Swap 0 blocks (0 = disabled) CPU Offload True Image Generation False No Block Prep False Note Models will be loaded in main process (Clear All Memory disabled) ================================================================================ Removing weight norm... [OK] OviFusionEngine initialized successfully (models will load on first generation) [PROMPT 11] Processing A man in a workshop, with soldering irons and circ... [GENERATION 11] Starting with seed 99 [GENERATION LORA] No LoRAs applied (using base model only) [DEBUG] No text_embeddings_cache available - T5 will be loaded in-process [SAGE ATTENTION] Enabled - using Sage Attention for ~10% speedup & lower VRAM [T5 CACHE] Cache key 820376c1545da46ef0063d280d415995337c1ff485c20951ef2635895a4dde38 [DEBUG] No cached embeddings - will load T5 ================================================================================ STEP 12 Loading T5 text encoder FIRST to minimize RAM usage ================================================================================ ================================================================================ Loading OVI models for first generation... Block Swap 0 blocks CPU Offload True ================================================================================ Initial VRAM 0.00 GB ================================================================================ SCALED FP8 T5 Loading T5 in Scaled FP8 format Expected VRAM savings ~50% (~5-6GB saved) ================================================================================ [FP8 CACHE] Found cached FP8 checkpoint COvi_Pro_v8Ovi_ProckptsWan2.2-TI2V-5Bmodels_t5_umt5-xxl-enc-fp8_scaled.safetensors [FP8 CACHE] Creating structure on CPU first (avoids BF16 VRAM allocation) [T5 LOAD][FP8] Structure created on CPU in 169.04s (FP8 cached path) WARNINGrootFailed to load cached FP8 checkpoint (The paging file is too small for this operation to complete. (os error 1455)). Rebuilding FP8 weights... [FP8 CACHE] Failed to load cached FP8 checkpoint (The paging file is too small for this operation to complete. (os error 1455)). Regenerating... [FP8 CACHE] No valid cache found - will quantize from scratch [SUBPROCESS] Generation failed with return code 3221225477 [GENERATION 11] No output file found in COvi_Pro_v8Ovi_Prooutputs after retries Total generation time 202.36 seconds ================================================================================ VIDEO GENERATION COMPLETED Final output path None File exists No ================================================================================ [MEMORY CLEANUP] Final cleanup completed - all generation memory freed

hello my friend. The latest 8 version seems to be doing better but running a test with the first T2V with Extention did not work properly. It completed the first video, did create a second video, the first video was not created and it fails to merge because the 3rd video is missing, but I notice it dropped out of the venv on my F drive which is why there app is, to the system python on the C drive. I will try to do another run.

Please start using Python 3.12, I don't understand these OLD Python versions; either that or maybe start using Conda to manage installations since you can easily control Python versions that way too.