Uploaded October 2024 | Updated September 2026, 3 hours ago
In our recent test, we explored two different approaches to AI-generated video content. The resulting video demonstrates the capabilities of both tools side by side.
Video Content
Left side: Deep fake using Deep Live Cam, where only the face is replaced
Right side: Completely AI-designed woman using Comfy UI
Deep Live Cam (Left Side)
Tutorial: youtube.com/watch?v=CIQ66fIAGBc
Process:
1. Install the tool
2. Upload a picture of the face to be used as replacement
3. Upload a target video where the new face will be applied
Strengths:
- Performs well with frontal camera angles
- Maintains convincing results with minimal facial movements
Limitations:
- Camera Angles: Non-frontal angles produced unrealistic results and visible artifacts
- Movement: Significant facial movements led to decreased quality
- Lighting: Shadow artifacts appeared on facial features, particularly on cheekbones
- Consistency: Occasional issues with mouth positioning, appearing slightly open when it shouldn't be
- Hair Rendering: Subtle flickering observed in hair across all tests
Comfy UI (Right Side)
Tutorial: youtube.com/watch?v=d3vpKqTiTvc
Process:
1. Install ComfyUI and download the AI model
2. Upload a video or photo
3. Let the tool create a pose figure from the movements in the video
4. Describe the new character you want to create
Pros and Cons:
- Useful for the first 80% of model creation, especially in client interactions
- Struggles with the final 20% of refinement
- Changing small details often affects other aspects of the model
- Quick and useful for initial concepts
- Not ideal for achieving 100% perfect results
This comparison highlights the strengths and limitations of two different approaches to AI-generated video content, showcasing the current state of deep fake and AI design technology.
#softmachine #ai #technology #innovation #deepfake
In our recent test, we explored two different approaches to AI-generated video content. The resulting video demonstrates the capabilities of both tools side by side.
Video Content
Left side: Deep fake using Deep Live Cam, where only the face is replaced
Right side: Completely AI-designed woman using Comfy UI
Deep Live Cam (Left Side)
Tutorial: youtube.com/watch?v=CIQ66fIAGBc
Process:
1. Install the tool
2. Upload a picture of the face to be used as replacement
3. Upload a target video where the new face will be applied
Strengths:
- Performs well with frontal camera angles
- Maintains convincing results with minimal facial movements
Limitations:
- Camera Angles: Non-frontal angles produced unrealistic results and visible artifacts
- Movement: Significant facial movements led to decreased quality
- Lighting: Shadow artifacts appeared on facial features, particularly on cheekbones
- Consistency: Occasional issues with mouth positioning, appearing slightly open when it shouldn't be
- Hair Rendering: Subtle flickering observed in hair across all tests
Comfy UI (Right Side)
Tutorial: youtube.com/watch?v=d3vpKqTiTvc
Process:
1. Install ComfyUI and download the AI model
2. Upload a video or photo
3. Let the tool create a pose figure from the movements in the video
4. Describe the new character you want to create
Pros and Cons:
- Useful for the first 80% of model creation, especially in client interactions
- Struggles with the final 20% of refinement
- Changing small details often affects other aspects of the model
- Quick and useful for initial concepts
- Not ideal for achieving 100% perfect results
This comparison highlights the strengths and limitations of two different approaches to AI-generated video content, showcasing the current state of deep fake and AI design technology.
#softmachine #ai #technology #innovation #deepfake
