Children’s Drawings to Animated Videos
Budget: €10,000 – €20,000 EUR
I have a working concept that turns a child's sketch into a colourful cartoon frame; now I need you to take those frames all the way to a polished 10-second (or longer) animated clip.
The finished video must maintain a traditional cartoon aesthetic with smooth, fluid motion and temporal consistency throughout.
Environment you will build:
- ComfyUI as the workflow engine (Diffusers models where applicable)
- SVD / AnimateDiff / CogVideoX for video generation and extension
- ControlNet and LoRA to preserve line work while adding natural motion
- Depth estimation feeding parallax/camera effects before final render
- Containerised in Docker, exposed via FastAPI, queued with Redis, optimised for VRAM-efficient multi-GPU deployment
Only open-source models with commercial-friendly licences.
Deliverables:
1. Fully automated pipeline (drawings in → MP4 out) as Docker images
2. FastAPI endpoints with example requests for each stage
3. Helm or Compose file for one-command GPU cluster deployment
4. Markdown documentation: setup, model weights, VRAM tips
5. Sample 10-second cartoon clip from a supplied doodle
Acceptance criteria:
- 24 fps playback
- Consistent character identity across frames
- Visible depth parallax effect
- Runs end-to-end on a single A100 40GB without OOM
Share repos or demos proving video-generation experience with the above stack, plus a rough milestone timeline.
The finished video must maintain a traditional cartoon aesthetic with smooth, fluid motion and temporal consistency throughout.
Environment you will build:
- ComfyUI as the workflow engine (Diffusers models where applicable)
- SVD / AnimateDiff / CogVideoX for video generation and extension
- ControlNet and LoRA to preserve line work while adding natural motion
- Depth estimation feeding parallax/camera effects before final render
- Containerised in Docker, exposed via FastAPI, queued with Redis, optimised for VRAM-efficient multi-GPU deployment
Only open-source models with commercial-friendly licences.
Deliverables:
1. Fully automated pipeline (drawings in → MP4 out) as Docker images
2. FastAPI endpoints with example requests for each stage
3. Helm or Compose file for one-command GPU cluster deployment
4. Markdown documentation: setup, model weights, VRAM tips
5. Sample 10-second cartoon clip from a supplied doodle
Acceptance criteria:
- 24 fps playback
- Consistent character identity across frames
- Visible depth parallax effect
- Runs end-to-end on a single A100 40GB without OOM
Share repos or demos proving video-generation experience with the above stack, plus a rough milestone timeline.