Uploaded November 2025 | Updated September 2026, 2 days ago
Replicate has over 50,000 models and it allows you to stitch together AI solutions. Cloudflare has a Developer Platform that allows you to stitch together primitives to build, host, and scale secure global web applications.
Merging the two together makes magic happen.
In this video Craig shows how he built a Voice based Image Editing app. He used Replicate for an image gen model, a image editing model, and an LLM. He hosted it on Workers and used the Agents SDK for state synchronnization and storage. He uses a WebSocket based speech to text model that provides turn detection that is hosted on Cloudflare Workers AI to produce an app that allows you to edit photos with your voice in real-time.
Repo: github.com/craigsdennis/speak-to-image-replicate-agents
## Models used
Flux Schnell Ultra: replicate.com/prunaai/flux-schnell-ultra
Qwen Image Edit Plus: replicate.com/qwen/qwen-image-edit-plus
Gemini Flash 2.5: replicate.com/google/gemini-2.5-flash
Deepgram Flux: developers.cloudflare.com/workers-ai/models/flux
Tools mentioned:
replicate.com
agents.cloudflare.com
developers.cloudflare.com
https://developers.cloudflare.comr/r2
developers.cloudflare.com/workers
developers.cloudflare.com/workers-ai
Chapters:
00:00 Intro
00:30 Demo: Generate an image
01:40 Demo: Edit an image
02:30 Demo: Edit the image with text to speech
04:20 Code: Server side API to create Agent
05:15 Code: Explore the Agent class and state
06:05 Code: React Front End Code with useAgent
07:29 Code: Deepgram Flux for end of turn detection
10:15 Code: Use a durable execution Workflow to save to long term object storage
12:20 Review key concepts
#cloudflare #cloudflareworkers #developer
Replicate has over 50,000 models and it allows you to stitch together AI solutions. Cloudflare has a Developer Platform that allows you to stitch together primitives to build, host, and scale secure global web applications.
Merging the two together makes magic happen.
In this video Craig shows how he built a Voice based Image Editing app. He used Replicate for an image gen model, a image editing model, and an LLM. He hosted it on Workers and used the Agents SDK for state synchronnization and storage. He uses a WebSocket based speech to text model that provides turn detection that is hosted on Cloudflare Workers AI to produce an app that allows you to edit photos with your voice in real-time.
Repo: github.com/craigsdennis/speak-to-image-replicate-agents
## Models used
Flux Schnell Ultra: replicate.com/prunaai/flux-schnell-ultra
Qwen Image Edit Plus: replicate.com/qwen/qwen-image-edit-plus
Gemini Flash 2.5: replicate.com/google/gemini-2.5-flash
Deepgram Flux: developers.cloudflare.com/workers-ai/models/flux
Tools mentioned:
replicate.com
agents.cloudflare.com
developers.cloudflare.com
https://developers.cloudflare.comr/r2
developers.cloudflare.com/workers
developers.cloudflare.com/workers-ai
Chapters:
00:00 Intro
00:30 Demo: Generate an image
01:40 Demo: Edit an image
02:30 Demo: Edit the image with text to speech
04:20 Code: Server side API to create Agent
05:15 Code: Explore the Agent class and state
06:05 Code: React Front End Code with useAgent
07:29 Code: Deepgram Flux for end of turn detection
10:15 Code: Use a durable execution Workflow to save to long term object storage
12:20 Review key concepts
#cloudflare #cloudflareworkers #developer
