Objective : Testing the text2video generation with LTX Video v0.9.5.
Check the original article at : https://blog.comfy.org/p/ltx-video-095-day-1-support-in-comfyui
For installation follow in the blog information.
Attached to the article you can find the original workflow ltxv_text_to_video_0.9.5.json
I have changed the Vae Decode to Vae Decode (Tiled)
It generates videos 768x512 with 97 frames in few seconds
Size of webp are more that 5MB and I cannot show the videos here.
The result for places looks quite real.
Changing the size to 608x416 the video it speed up the generation to ~24 seconds
Quality is not the best. To try this workflow use the LTX_I2V_Step1.json
I also did a test 1504x1024 size and the quality was ok
Check the ltx__videos.zip file to see the example here.
Last Test:
I want to generate a video with this prompt:
A traditional wooden Japanese boat with intricate carvings glides across a calm lake in Japan, with gentle ripples in the water. A lone samurai stands on the boat, wearing a black and red armor set with gold detailing, a menacing yet elegant kabuto (helmet) adorned with a crescent moon ornament. His katana rests at his side, partially unsheathed, reflecting the warm hues of the setting sun. The background features the majestic Mount Fuji, its snow-capped peak standing tall against a sky painted with vibrant shades of orange, pink, and purple from the evening sun. Cherry blossom trees line the distant shore, their delicate pink petals occasionally drifting onto the water’s surface. Mist rolls over the lake, giving the scene a mystical, almost ethereal atmosphere. The boat is lit by traditional paper lanterns hanging from wooden poles, casting a soft, golden glow that reflects off the rippling water. The samurai's face is partially shadowed, showing a deep, contemplative expression as he gazes towards the horizon. The composition is balanced with the samurai and boat in the foreground, Mount Fuji in the background, and the tranquil lake stretching between them. The overall color palette consists of deep blues, purples, and oranges, creating a harmonious blend of warmth and serenity. The image captures a cinematic, highly detailed, and immersive atmosphere, evoking a sense of honor, tradition, and the beauty of Japan’s natural landscape.I changed the length of the video to 177 frames (check the file : LTX_I2V_Step2.json )
I run a few video generations (the quality is not really good), file Videos_mp4.zip
The quality of the videos is not really good.
Conclusion: I won't use anymore this model. Wan2.1 and Hunyuan I2V have a really good quality compared to this one.