Image to Talking Video · Image-to-Video
Upload a close-up photo and turn it into a short video clip with lip-synced speech and stereo audio, powered by Flux 2 Klein and MiniMax H3. Pick a reaction, describe your subject, and hit run.
flux 2 klein
image to video
minimax h3
27
ABOUT THE WORKFLOW
Turn a photo into a talking video clip
Upload a close-up portrait, pick one of five pre-written reaction scenes, and get back a short video of that person speaking with synced lip movement and background audio. The workflow generates a new still frame from your photo, then animates it into a finished video with dialogue. You get both the still image and the final video file.
Model
MiniMax H3 by MiniMax (Beijing), also known as Hailuo 3.0. Strong at cinematic talking-head clips, lip-synced dialogue, and reference-driven video with consistent identity. The still frame is built by Flux 2 Klein 9B from Black Forest Labs.
HOW IT WORKS
Step 1. Upload a close-up photo
Upload a portrait-style photo of a person. Tight framing from upper chest to above the head works best. The workflow preserves the face, clothing, and pose from your photo.
Works great with: headshots · editorial portraits · lifestyle close-ups
Step 2. Pick a reaction number (1 to 5)
Type a number from 1 to 5. Each number selects a different reaction scene with its own camera direction, dialogue line, and audio cues. For example, reaction 1 produces “Oh my God... did you hear that? That was insane!” while reaction 3 gives “No way... tell me you heard that!”
Step 3. Describe the subject (English)
Write a detailed description of the person and scene in your photo. Include face, hair, clothing, pose, and background details. The closer this matches your photo, the better the generated frame preserves identity.
Example: “A young woman with short red hair, wearing a denim jacket, standing in front of a coffee shop with warm lighting”
Step 4. Set duration (optional)
Choose the video length in seconds. Default is 6. The workflow rounds to the nearest valid frame count automatically.
Step 5. Run and download
Hit run. Flux 2 Klein generates a new still frame matching the chosen reaction, then MiniMax H3 animates it into a video with lip-synced speech and stereo audio. You get a still image and a finished video file.
Ready for: social media · presentations · product demos
First time? Leave every setting as-is. The defaults are the right starting point for almost everyone.
RECOMMENDED SETTINGS
Quick-start guide. Find the goal that matches yours and copy the settings.
Quick test run — Reaction number: 1, Duration: 6 seconds, all other defaults. The right starting point for almost everyone.
Want a different dialogue line? — Change the reaction number. Each number (1 through 5) uses a different spoken line and camera direction. Try a few to find the tone you need.
Want a longer clip? — Set duration to 8 or 10 seconds. Quality can drop on clips longer than 10 seconds, so stay in the 4 to 8 second range for the best results.
Need a sharper still frame? — Increase Image Steps from 20 to 30. More passes add detail to the generated frame before video animation begins.
Face doesn’t match the photo? — Rewrite the subject description with more detail. Include hair color and texture, skin tone, clothing colors, patterns, accessories, and exact pose. The description is what anchors identity preservation.
Video looks distorted or drifts? — Lower duration to 4 seconds and re-run. Shorter clips hold identity and anatomy better. Also check that your subject description closely matches the uploaded photo.
Prompt: Write the subject description as if you’re telling someone exactly what your photo looks like. Be specific about hair, face, clothing, and pose. “A woman with curly brown hair and a green sweater, sitting at a desk” works. “A person” does not.
LEARN
📹 Videos
ComfyUI 101 Free Course ft. Sebastian Kamph
Floyo 101 for Team Collaboration
✨ Quick links
USE CASES
🎬 Social media reaction clips
Upload a headshot and generate a short reaction video for TikTok, Reels, or YouTube Shorts without filming anything.
📢 Product reveal moments
Create a talking-head clip of someone reacting to a product launch, announcement, or surprise reveal for marketing content.
🎓 Training and presentation inserts
Generate a quick talking clip to break up slide decks and e-learning modules with a human presence.
🎭 Character prototyping for creative projects
Test how a character looks and sounds on camera before committing to a full production shoot.
WHAT WORKS BEST / WHAT TO AVOID
✅ Works great
Tight close-up photos with clear face and upper body visible
Well-lit portraits with a single subject
Detailed subject descriptions that match the uploaded photo closely
Durations between 4 and 8 seconds
⚠️ May produce softer results
Wide shots or group photos with multiple people visible
Low-resolution or heavily compressed source images
Vague subject descriptions with missing clothing or hair details
Durations longer than 10 seconds, where identity and anatomy can drift
FAQ
What is MiniMax H3?
MiniMax H3 is a video generation model by MiniMax (Beijing), also known as Hailuo 3.0. It takes a first-frame image and a text prompt with scene directions, dialogue, camera, and audio cues. Output is up to 15 seconds of video at 1024x576, 24 fps, with native 32 kHz stereo audio including speech.
How does the image to talking video workflow work?
The workflow runs two models back to back. First, Flux 2 Klein 9B generates a new still frame from your uploaded photo and the selected reaction scene. Then MiniMax H3 animates that frame into a video with lip-synced dialogue, background audio, and camera movement. You get both the still frame and the final video.
Can I use MiniMax H3 for commercial projects?
MiniMax H3 uses the MiniMax H3 Community License, which allows open use. However, Flux 2 Klein uses the FLUX Non-Commercial License for the still-frame generation step. Check both licenses before using outputs in commercial work.
What resolution does the output video use?
Output resolution is fixed at 1024x576 (16:9 widescreen) at 24 fps. The workflow handles frame count rounding automatically from your duration value.
Does MiniMax H3 support custom dialogue?
This workflow includes five pre-written reaction scenes with fixed dialogue lines. Each reaction number selects a different spoken line, camera direction, and audio treatment. To use fully custom dialogue, you would need to modify the prompt banks in the workflow.
Is MiniMax H3 good for long videos?
MiniMax H3 handles clips up to about 15 seconds, but quality is strongest in the 4 to 8 second range. Longer clips can show identity drift, hand distortion, or anatomy issues. Start short and increase duration only if you need it.
How to run MiniMax H3 image to video online?
You can run MiniMax H3 image to video online through Floyo. No installation, no setup, no API key to wire up. Open the workflow in your browser, upload your inputs, and hit run. Free to try.
WHY FLOYO?
Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.
A designer runs an edit and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.
For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.
Ready to try it?
Upload a close-up photo, pick a reaction, and generate your first talking video clip.
Questions? Watch the free course or check the FAQ above.
Read more


