Bring in a document.
Keep the document on the canvas and reuse it.
Modality-First
Connect. Create.
Connect text, images, audio and video into your own AI workflow.
Open-source core · Your models · Self-hostable
Text
Image
Audio
3D
Video
One workflow / Four steps
Start with a document.
Extract the text.
Extract the text for the next step.
Branch in two directions.
Turn the same text into a summary and speech.
Connect an image.
Combine the audio and a character image into a talking video.
Workflow illustration · Requires compatible pluginsScroll to build the workflow
Think in possibilities.
Build with modalities.
Change the model without rebuilding the workflow.
Why TongFlow starts with modalities →What would you start with?
Start here Transform Continue from here
Text
Develop an idea
Text
Generate an image
Image
Give it a voice
Audio
Start here Transform Continue from here
Image
Understand the image
Text
Set it in motion
Video
Build in 3D
3D
Start here Transform Continue from here
Audio
Transcribe speech
Text
Reduce noise
Audio
Separate sounds
Audio
Start here Transform Continue from here
Video
Understand a clip
Text
Extract a frame
Image
Extract the audio
Audio
Capability examples. Running requires compatible plugins.
Operations
Add. Transform. Combine. Split.
Add
Start with text, a file or a link.
Transform
Turn an image into text, video or another image.
Combine
Combine an image and a voice into video.
Split & batch
Split into parts and process each one.
Scroll to follow the connections
Connect the steps your task needs.
Make information easier to use
Document Text Summary Audio
Take an image further
Image Transform Video / 3D
Find the useful parts
Video Clips Audio Text
MODALITY-FIRST
Materials, capabilities and models.
Every result is a new beginning
Connect uploaded and generated materials to the next step.
Choose a capability. Then a model.
Switch compatible models while keeping your workflow.
A shared structure for people and agents
Build on the canvas or let an agent create an editable workflow.
Scroll to follow the connections
One image. One motion reference.
Combine both inputs with a compatible motion-transfer plugin.
Watch the example
Cloud, self-hosted, or built with code.
One open-source core. Connect the plugins and providers you need.
TongFlow Cloud
Connect your API keys or Modal account to run.
Open the workspaceSelf-host
Configure plugins and models on your own infrastructure.
Get the sourceBuild with an agent
Build with TongFlow or DSH; edit on the canvas.
Explore the DSH pluginGet the macOS / Windows app ↗The desktop app opens the cloud workspace; it requires a connection.
A few things to know
What is TongFlow?
TongFlow is an open-source tool for building multimodal AI workflows. Connect text, images, audio, video, 3D and documents on a visual canvas to extract, transform, generate or combine information. Each step runs through a compatible plugin and its model or processing implementation. You can use the cloud workspace or self-host the app.
What does modality-first mean?
You build around forms of information, such as text, images or sound. Capabilities connect those forms; models implement the capabilities. This gives different workflows a common structure.
Can I connect anything to anything?
Connections follow the input and output types of each capability. A workflow also needs compatible plugins. The design is open-ended, but available operations determine what can run today.
Does it only generate media?
No. Understanding an image, extracting text, transcribing speech, splitting a video and combining materials are also part of the system. Generation is one kind of operation.
What do I need to run a workflow?
Use the cloud workspace or self-host the app, then configure the provider credentials required by your chosen plugins. Modal-backed plugins require a Modal account; API plugins require the relevant provider keys.
How does billing work?
Cloud access is a subscription. Model API usage and GPU compute are billed separately by your providers. TongFlow does not sell generation credits. See pricing for the current subscription and trial details.
