docs: add MiniMax H3 API workflow tutorial under partner nodes - #1343
Conversation
|
Warning Review limit reached
Next review available in: 49 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Organization UI Review profile: ASSERTIVE Plan: Pro Plus Run ID: 📒 Files selected for processing (4)
📝 WalkthroughWalkthroughMiniMax H3 tutorials were added in English, Chinese, Japanese, and Korean. Each page documents cloud text-to-video, first/last-frame, and reference-to-video workflows. Navigation entries link to each localized tutorial. ChangesMiniMax H3 tutorial updates
Possibly related PRs
🚥 Pre-merge checks | ✅ 2✅ Passed checks (2 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
✨ Simplify code
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tutorials/video/minimax/minimax-h3.mdx`:
- Around line 253-255: Update the API download card search terms to distinguish
them from native workflows: use "MiniMax H3 T2V (API)" at
tutorials/video/minimax/minimax-h3.mdx lines 253-255 and the corresponding
Japanese, Korean, and Chinese sites at lines 264-266, 267-269, and 265-267; use
"MiniMax H3 R2V (API)" at tutorials/video/minimax/minimax-h3.mdx lines 310-312
and the corresponding Japanese, Korean, and Chinese sites at lines 321-323,
324-326, and 322-324.
- Around line 300-304: Update the MiniMax H3 API R2V section under “MiniMax H3
Reference to Video (API)” to use the workflow’s video preview asset instead of
the static .webp image, matching the API template’s available output. Align the
API prompt examples and R2V guidance with api_minimax_h3_r2v.json by using the
Image 1 reference tag consistently, while leaving the native R2V Picture 1
syntax unchanged.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro Plus
Run ID: b12185c4-72d5-4324-bf22-36bdf1a20e7a
📒 Files selected for processing (4)
ja/tutorials/video/minimax/minimax-h3.mdxko/tutorials/video/minimax/minimax-h3.mdxtutorials/video/minimax/minimax-h3.mdxzh/tutorials/video/minimax/minimax-h3.mdx
| <Card title="Download Workflow" icon="download" href="https://github.com/Comfy-Org/workflow_templates/blob/main/templates/api_minimax_h3_t2v.json"> | ||
| Download JSON or search "MiniMax H3 T2V" in Template Library | ||
| </Card> |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
API download cards reuse the native workflow's search term. All eight sites tell users to search the Template Library using the exact string already used by the corresponding native (open-weights) workflow. FLF2V-API avoids this by using the distinct term "MiniMax H3 FLF2V", since no native FLF2V template shares that name. T2V-API and R2V-API don't get the same treatment, so a user searching "MiniMax H3 T2V" or "MiniMax H3 R2V" cannot tell from the term alone which result is the API version. One extra word closes this gap: add "(API)" to the string, and the puzzle of "which one do I click" resolves itself.
tutorials/video/minimax/minimax-h3.mdx#L253-L255: change the search term to"MiniMax H3 T2V (API)".tutorials/video/minimax/minimax-h3.mdx#L310-L312: change the search term to"MiniMax H3 R2V (API)".ja/tutorials/video/minimax/minimax-h3.mdx#L264-L266: change the search term to"MiniMax H3 T2V (API)".ja/tutorials/video/minimax/minimax-h3.mdx#L321-L323: change the search term to"MiniMax H3 R2V (API)".ko/tutorials/video/minimax/minimax-h3.mdx#L267-L269: change the search term to"MiniMax H3 T2V (API)".ko/tutorials/video/minimax/minimax-h3.mdx#L324-L326: change the search term to"MiniMax H3 R2V (API)".zh/tutorials/video/minimax/minimax-h3.mdx#L265-L267: change the search term to"MiniMax H3 T2V (API)".zh/tutorials/video/minimax/minimax-h3.mdx#L322-L324: change the search term to"MiniMax H3 R2V (API)".
📍 Affects 4 files
tutorials/video/minimax/minimax-h3.mdx#L253-L255(this comment)tutorials/video/minimax/minimax-h3.mdx#L310-L312ja/tutorials/video/minimax/minimax-h3.mdx#L264-L266ja/tutorials/video/minimax/minimax-h3.mdx#L321-L323ko/tutorials/video/minimax/minimax-h3.mdx#L267-L269ko/tutorials/video/minimax/minimax-h3.mdx#L324-L326zh/tutorials/video/minimax/minimax-h3.mdx#L265-L267zh/tutorials/video/minimax/minimax-h3.mdx#L322-L324
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@tutorials/video/minimax/minimax-h3.mdx` around lines 253 - 255, Update the
API download card search terms to distinguish them from native workflows: use
"MiniMax H3 T2V (API)" at tutorials/video/minimax/minimax-h3.mdx lines 253-255
and the corresponding Japanese, Korean, and Chinese sites at lines 264-266,
267-269, and 265-267; use "MiniMax H3 R2V (API)" at
tutorials/video/minimax/minimax-h3.mdx lines 310-312 and the corresponding
Japanese, Korean, and Chinese sites at lines 321-323, 324-326, and 322-324.
| ### MiniMax H3 Reference to Video (API) | ||
|
|
||
| Generate videos conditioned on reference images, videos, and audio through the MiniMax H3 API. | ||
|
|
||
|  |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Check API R2V template default prompt tag format and preview video availability.
set -euo pipefail
echo "== Default widget/prompt values in api_minimax_h3_r2v.json =="
curl -s https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.json | grep -oE '"(Image|Picture|Video|Audio) [0-9]+' | sort -u
echo "== Checking for a preview video asset =="
curl -s -o /dev/null -w "HTTP %{http_code}\n" https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.mp4Repository: Comfy-Org/docs
Length of output: 214
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "== Locate minimax docs and workflow templates files =="
fd -a 'minimax-h3\.mdx|api_minimax_h3_r2v' . | sed 's#^\./##' | head -100
echo
echo "== MDX lines 280-330 =="
file="$(fd 'minimax-h3\.mdx' . | head -1)"
if [ -n "${file:-}" ]; then
nl -ba "$file" | sed -n '280,330p'
fi
echo
echo "== Nearby API preview snippets for T2V/FLF2V =="
rg -n 'api_|<video|\.webp|Reference to Video' "$file" -C 2 | sed -n '1,220p'
echo
echo "== Check exact remote API R2V json prompt/tags and output files =="
python3 - <<'PY'
import json, urllib.request, urllib.error
url = 'https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.json'
try:
data = json.load(urllib.request.urlopen(url, timeout=20))
except Exception as e:
print(f'remote json: {e}')
raise
print(type(data))
for key in ['workflow', 'inputs', 'models', 'node', 'nodes', 'nodes_properties']:
if key in data:
print(key, type(data[key]))
# Try likely prompt/input fields.
text=' '.join(str(data).split())
for pat in ['Image ', 'Picture ', 'Video ', 'Audio ']:
if pat in text:
idx = text.index(pat)
print(f'contains {pat!r} ... {text[max(0,idx-80):idx+80]}')
PY
echo
echo "== Check remote preview video URLs =="
for url in \
https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.mp4 \
https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.webm \
https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.mp4 \
https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.webm \
https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.webp \
https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.json; do
printf '%s ' "$url "
curl -s -o /dev/null -w '%{http_code}' --max-time 10 "$url" || printf 'ERR'
printf '\n'
done
echo
echo "== Existing preview video assets for MiniMax API workflows in template output/templates =="
curl -s https://api.github.com/repos/Comfy-Org/workflow_templates/contents/output?ref=main \
| python3 - <<'PY'
import json, sys, re
data=json.load(sys.stdin)
for item in data:
if item.get('type') == 'file' and 'minimax' in item.get('name','').lower() and re.search(r'\.(mp4|webm|webp|png)$', item.get('name','')):
print(item.get('download_url'))
PY
PYRepository: Comfy-Org/docs
Length of output: 514
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "== MDX lines 280-330 =="
nl -ba /home/jailuser/git/tutorials/video/minimax/minimax-h3.mdx | sed -n '280,330p'
echo
echo "== Nearby API preview snippets for T2V/FLF2V =="
rg -n 'api_|<video|\.webp|Reference to Video|Image 1|Picture 1|Video 1|Audio 1' /home/jailuser/git/tutorials/video/minimax/minimax-h3.mdx -C 2
echo
echo "== Check exact remote API R2V json prompt/tags and output files =="
python3 - <<'PY'
import json, sys, urllib.request, urllib.error, re
url = 'https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.json'
try:
data = json.load(urllib.request.urlopen(url, timeout=20))
except Exception as e:
print(f'remote json: {e}')
raise
def walk(obj, depth=0):
if depth > 6:
return
if isinstance(obj, dict):
for k, v in obj.items():
if isinstance(v, (str, int, float)) and re.search(r'(?:Image|Picture|Video|Audio)\s+[0-9]+?', str(v)):
print(k, "=", repr(v))
elif isinstance(v, (dict, list)) and k not in ['base64', 'data']:
walk(v, depth+1)
elif isinstance(obj, list):
for i, v in enumerate(obj[:200]):
walk(v, depth+1)
walk(data)
for path, ext in [
('main/output', 'mp4'), ('main/output', 'webm'),
('main/templates', 'mp4'), ('main/templates', 'webm'), ('main/templates', 'webp'),
]:
full = f'https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.{ext}' if path == 'main/output' else f'https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.{ext}'
try:
r = urllib.request.urlopen(full, timeout=10)
print(full, r.getcode(), r.headers.get('Content-Type'))
except urllib.error.HTTPError as e:
print(full, e.code, e.reason)
except Exception as e:
print(full, 'ERR', type(e).__name__, str(e))
PYRepository: Comfy-Org/docs
Length of output: 216
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "== MDX file stats and target lines =="
wc -l /home/jailuser/git/tutorials/video/minimax/minimax-h3.mdx
sed -n '280,330p' /home/jailuser/git/tutorials/video/minimax/minimax-h3.mdx
sed -n '210,230p' /home/jailuser/git/tutorials/video/minimax/minimax-h3.mdx
sed -n '320,328p' /home/jailuser/git/tutorials/video/minimax/minimax-h3.mdx
echo
echo "== Nearby API preview snippets for MiniMax native/API =="
rg -n 'api_|<video|\.webp|Reference to Video|Image 1|Picture 1|Video 1|Audio 1' /home/jailuser/git/tutorials/video/minimax/minimax-h3.mdx -C 2
echo
echo "== Check exact remote API R2V json prompt/tags and possible preview assets =="
python3 - <<'PY'
import json, sys, urllib.request, urllib.error, re
url = 'https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.json'
data = json.load(urllib.request.urlopen(url, timeout=20))
def walk(obj, depth=0):
if depth > 6:
return
if isinstance(obj, dict):
for k, v in obj.items():
if k in ("prompt", "positive", "negative", "reference", "inputs") and re.search(r'(?i)(?:Image|Picture|Video|Audio)\s*[0-9]+?', str(v)):
print(f"{k} =", repr(v))
elif isinstance(v, (dict, list)) and k not in ["base64", "data"]:
walk(v, depth+1)
elif isinstance(obj, list):
for item in obj:
walk(item, depth+1)
walk(data)
for path, ext in [
("main/output", "mp4"), ("main/output", "webm"),
("main/templates", "mp4"), ("main/templates", "webm"), ("main/templates", "webp"),
]:
prefix = path if path == "main/output" else "main/templates"
full = f"https://raw.githubusercontent.com/Comfy-Org/workflow_templates/{prefix}/api_minimax_h3_r2v.{ext}"
try:
r = urllib.request.urlopen(full, timeout=10)
print(full, r.getcode(), r.headers.get("Content-Type"))
except urllib.error.HTTPError as e:
print(full, e.code, e.reason)
except Exception as e:
print(full, "ERR", type(e).__name__, str(e))
PYRepository: Comfy-Org/docs
Length of output: 12017
🌐 Web query:
Comfy-Org workflow_templates api_minimax_h3_r2v.json prompt tags Image Picture Video Audio
💡 Result:
The api_minimax_h3_r2v.json file is a template provided by the Comfy-Org/workflow_templates repository, designed for use with the MiniMax H3 multimodal generation model [1][2]. This template specifically leverages the "Reference-to-Video" (R2V) capabilities of the H3 model [3][4][5]. Unlike standard Text-to-Video (T2V) or simple Image-to-Video (I2V) workflows, the R2V mode allows for complex, multi-modal context input [3][6]. Key details regarding the api_minimax_h3_r2v.json workflow: Modality and Prompting: The H3 model architecture treats text, image, video, and audio as unified context [6]. In an R2V workflow, your prompt functions as the central instruction, describing the desired scene, motion, camera behavior, and the relationship between your referenced assets [3][5][6]. You are encouraged to explicitly reference your inputs in the prompt (e.g., "Reference the motion from Video 1, character identity from Image 2, and audio rhythm from Audio 3") [5][6]. Supported Input Types and Limits (Omni Reference): The workflow supports an "Omni Reference" approach, allowing for the following input assets [5]: * Image: Up to 9 reference images (for identity, style, composition) [3][4][5]. * Video: Up to 3 video clips (for motion, pacing, camera behavior) [3][4][5]. * Audio: Up to 3 audio clips (for voice, ambiance, or timing) [3][4][5]. General Constraints: * Total Assets: Typically capped at 12 assets total [5]. * Clip Duration: Video and audio clips are generally restricted to 2–15 seconds, with a 15-second total budget per media type [3][5]. * Audio Usage: Audio cannot be submitted alone in R2V mode; it must be accompanied by at least one image or video reference [3][4][5]. * Resolution/Output: The model natively supports up to 2K output with native stereo sound [6]. Technical Implementation: The workflow utilizes ComfyUI's API node system [7][8]. The JSON file is structured in the standard ComfyUI API format—which excludes UI layout data (positions, colors) and relies on numeric node IDs—to ensure it is optimized for programmatic API submission to the MiniMax service [9]. You can find this and other official templates in the Comfy-Org/workflow_templates repository [1].
Citations:
- 1: https://github.com/comfy-org/workflow_templates
- 2: Update template to 0.1.65 ComfyUI#9501
- 3: https://reapi.ai/docs/minimax-h3
- 4: https://www.runcomfy.com/models/minimax/minimax-h3/reference-to-video
- 5: https://dev.to/agi2y/designing-a-practical-minimax-h3-video-workflow-text-frames-and-omni-references-2747
- 6: https://www.minimax.io/blog/minimax-h3
- 7: https://github.com/Comfy-Org/ComfyUI/blob/5e3f15a8/comfy_api_nodes/nodes_minimax.py
- 8: https://docs.comfy.org/zh-CN/built-in-nodes/partner-node/video/minimax/minimax-text-to-video
- 9: https://docs.comfy.org/development/api-development/workflow-api-format
🏁 Script executed:
#!/bin/bash
set -euo pipefail
python3 - <<'PY'
import json, urllib.request, urllib.error
for url, insecure in [
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.json", False),
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v-1.webp", False),
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.mp4", False),
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.webm", False),
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v.json", True),
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/templates/api_minimax_h3_r2v-1.webp", True),
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.mp4", True),
("https://raw.githubusercontent.com/Comfy-Org/workflow_templates/main/output/api_minimax_h3_r2v.webm", True),
]:
print(f"-- {url} insecure={insecure} --")
opener = urllib.request.build_opener()
if insecure:
import ssl
opener.add_handler(urllib.request.HTTPSHandler(context=ssl._create_unverified_context()))
try:
with opener.open(url, timeout=15) as r:
body = r.read(65536)
print(r.status, r.headers.get("Content-Type"))
try:
data = json.loads(body)
print("is JSON")
print(data.get("class_type"))
for k in ("prompt", "positive", "negative", "reference_image", "reference_video", "reference_audio", "first_frame", "last_frame"):
if k in data.get("inputs", {}):
print(f"{k} =", repr(data["inputs"][k][:200] if isinstance(data["inputs"][k], str) else data["inputs"][k]))
except json.JSONDecodeError:
print("is non-JSON byte image", body[:5])
except urllib.error.HTTPError as e:
print("HTTP", e.code, e.reason)
except Exception as e:
print("ERR", type(e).__name__, str(e))
PYRepository: Comfy-Org/docs
Length of output: 2306
🌐 Web query:
github.com Comfy-Org workflow_templates api_minimax_h3_r2v.json download
💡 Result:
The file api_minimax_h3_r2v.json is part of the official ComfyUI workflow templates maintained in the Comfy-Org/workflow_templates repository [1][2]. These templates are managed as part of the comfyui-workflow-templates dependency package, which provides native model workflow support within the ComfyUI interface [3][4][5]. To download or access this file, you should visit the official GitHub repository at: https://github.com/Comfy-Org/workflow_templates Once in the repository, you can locate the file within the templates directory: /templates/api_minimax_h3_r2v.json The repository is structured to hold template JSON files and their associated metadata [1][6]. You can download the file directly through the GitHub web interface by navigating to the file path and clicking the download or raw button, or by cloning the repository to your local machine using git: git clone https://github.com/Comfy-Org/workflow_templates.git Please note that these templates are designed to be used within the ComfyUI application itself, which has a dedicated template browser accessible via the menu (Workflow -> Browse Workflow Templates) [3][5]. While you can manually download the JSON, the recommended way to access and use these workflows is through the built-in ComfyUI template system, which automatically handles dependencies and updates [3][5].
Citations:
- 1: https://github.com/Comfy-Org/workflow_templates
- 2: https://github.com/comfy-org/workflow_templates
- 3: https://docs.comfy.org/interface/features/template
- 4: https://pypi.org/project/comfyui-workflow-templates/0.11.7/
- 5: https://comfyui-wiki.com/en/interface/features/template
- 6: https://github.com/Comfy-Org/workflow_templates/blob/main/docs/BLUEPRINTS.md
🌐 Web query:
"api_minimax_h3_r2v.json" "prompt" "Video 1" "Image 1"
💡 Result:
In the context of the MiniMax H3 video generation API, references to "Video 1" and "Image 1" within the prompt are used as explicit identifiers to instruct the model on how to utilize provided input assets [1][2][3][4]. When you supply multiple images, videos, or audio clips as inputs for "Reference-to-Video" (R2V) generation, the API associates them with these indices based on their order in the input list [2][4]. For example, the first image uploaded in your reference list is referred to as "Image 1," the second as "Image 2," and so on; similarly, the first video clip provided is "Video 1" [2][4]. To achieve high-quality results, it is a recommended best practice to use the prompt to explicitly define the role of each referenced asset [3][4]. Instead of merely providing the files, you should describe which asset should be used for specific tasks such as character identity, motion, style, or background [3][4]. Examples of effective prompting include: - "Use Image 1 as the character reference and Video 1 to drive the motion." [4] - "Use Image 1 for the overall mood, location, and film texture; Image 2 as the protagonist reference." [3] By explicitly naming the source and its intended function (e.g., "Image 1 for identity," "Video 1 for motion"), you help the MiniMax H3 model accurately apply the specific characteristics of your provided media to the generated output [3][4].
Citations:
- 1: https://reapi.ai/docs/minimax-h3
- 2: https://fal.ai/models/minimax/h3/reference-to-video/api
- 3: https://fal.ai/learn/devs/minimax-h3-prompting-guide
- 4: https://runware.ai/docs/models/minimax-h3/guides/motion-camera-performance
Use the API R2V video preview and keep the API tag syntax aligned with api_minimax_h3_r2v.json.
The surrounding MiniMax API sections use <video> previews, but line 304 uses the static .webp asset. If the workflow output video is ready, replace it here so the API section has the same “motion” preview promise as T2V and FLF2V.
The API prompt examples and the API R2V tip say Image 1, while the native R2V tip uses <Picture 1>; follow the API template’s reference labels so users do not bind prompts with the wrong tag names.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@tutorials/video/minimax/minimax-h3.mdx` around lines 300 - 304, Update the
MiniMax H3 API R2V section under “MiniMax H3 Reference to Video (API)” to use
the workflow’s video preview asset instead of the static .webp image, matching
the API template’s available output. Align the API prompt examples and R2V
guidance with api_minimax_h3_r2v.json by using the Image 1 reference tag
consistently, while leaving the native R2V Picture 1 syntax unchanged.
05260eb to
762deed
Compare
lin-bot23
changed the title
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tutorials/partner-nodes/minimax/minimax-h3-api.mdx`:
- Line 10: In the MiniMax H3 API description, update the verb in “no model
downloads or local GPU are required” to agree with the nearest singular noun
GPU, leaving the rest of the sentence unchanged.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro Plus
Run ID: cda3f5fd-3446-4cbe-b4b2-373365a6e05e
📒 Files selected for processing (5)
docs.jsonja/tutorials/partner-nodes/minimax/minimax-h3-api.mdxko/tutorials/partner-nodes/minimax/minimax-h3-api.mdxtutorials/partner-nodes/minimax/minimax-h3-api.mdxzh/tutorials/partner-nodes/minimax/minimax-h3-api.mdx
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tutorials/partner-nodes/minimax/minimax-h3.mdx`:
- Line 12: Update both duration references in
tutorials/partner-nodes/minimax/minimax-h3.mdx at lines 12-12 and 43-43 from
5-15 seconds to 4-15 seconds, preserving the surrounding overview and
text-to-video guidance.
- Line 75: Update the “Image constraints” item in the minimax-h3 tutorial to
document that each frame’s width and height must not exceed 5760 pixels, while
preserving the existing minimum dimensions and aspect-ratio requirements.
- Line 102: Update the “Limits” item in the minimax-h3 documentation to state
that mixed reference inputs are capped at 12 files total, while retaining the
existing per-type limits and duration constraints.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro Plus
Run ID: 5753a683-f831-4f78-ba73-ee67559234db
📒 Files selected for processing (5)
docs.jsonja/tutorials/partner-nodes/minimax/minimax-h3.mdxko/tutorials/partner-nodes/minimax/minimax-h3.mdxtutorials/partner-nodes/minimax/minimax-h3.mdxzh/tutorials/partner-nodes/minimax/minimax-h3.mdx
|
|
||
| [MiniMax H3](https://www.minimax.io/blog/minimax-h3) is MiniMax's general-purpose, omni-modal generation model, available in ComfyUI through the MiniMax H3 API nodes. The API workflows run generation on MiniMax's servers, so no model downloads or local GPU are required, and each second of generated video is billed to your Comfy API account. | ||
|
|
||
| MiniMax H3 generates video with **native stereo audio**: voice, sound effects, and music are modeled together in a single forward pass instead of being layered on afterward. Output is up to 2K resolution at 5-15 seconds per clip. |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Use the API duration range in both locations.
Both statements use 5-15 seconds, but MiniMax H3 accepts integer durations from 4 through 15 seconds. Update both references to 4-15 seconds. (platform.minimax.io)
tutorials/partner-nodes/minimax/minimax-h3.mdx#L12-L12: update the overview duration range.tutorials/partner-nodes/minimax/minimax-h3.mdx#L43-L43: update the text-to-video duration guidance.
📍 Affects 1 file
tutorials/partner-nodes/minimax/minimax-h3.mdx#L12-L12(this comment)tutorials/partner-nodes/minimax/minimax-h3.mdx#L43-L43
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@tutorials/partner-nodes/minimax/minimax-h3.mdx` at line 12, Update both
duration references in tutorials/partner-nodes/minimax/minimax-h3.mdx at lines
12-12 and 43-43 from 5-15 seconds to 4-15 seconds, preserving the surrounding
overview and text-to-video guidance.
| #### Prompting tips | ||
|
|
||
| 1. **Reference by tag**: Refer to each input by its order in the prompt, for example `Image 1`, `Video 1`, or `Audio 1` | ||
| 2. **Limits**: Up to 9 reference images, 3 reference videos, and 3 reference audio clips. Each video or audio clip must be 2-15 seconds; total reference video and audio is capped at 15 seconds each |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick win
Document the 12-file mixed-input cap.
The listed per-type limits total 15 assets, but a mixed reference request is capped at 12 files. A reader can follow this line and submit a request that the API rejects. Add the total cap. (platform.minimax.io)
Suggested text
-2. **Limits**: Up to 9 reference images, 3 reference videos, and 3 reference audio clips. Each video or audio clip must be 2-15 seconds; total reference video and audio is capped at 15 seconds each
+2. **Limits**: Up to 9 reference images, 3 reference videos, and 3 reference audio clips, with a maximum of 12 reference files per request. Each video or audio clip must be 2-15 seconds; total duration is capped at 15 seconds for videos and 15 seconds for audio📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| 2. **Limits**: Up to 9 reference images, 3 reference videos, and 3 reference audio clips. Each video or audio clip must be 2-15 seconds; total reference video and audio is capped at 15 seconds each | |
| 2. **Limits**: Up to 9 reference images, 3 reference videos, and 3 reference audio clips, with a maximum of 12 reference files per request. Each video or audio clip must be 2-15 seconds; total duration is capped at 15 seconds for videos and 15 seconds for audio |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@tutorials/partner-nodes/minimax/minimax-h3.mdx` at line 102, Update the
“Limits” item in the minimax-h3 documentation to state that mixed reference
inputs are capped at 12 files total, while retaining the existing per-type
limits and duration constraints.
2 participants
Add a new Partner Nodes tutorial page for the MiniMax H3 API workflows (
tutorials/partner-nodes/minimax/minimax-h3-api.mdx), covering the threeapi_minimax_h3_*templates:api_minimax_h3_t2v)api_minimax_h3_flf2v)api_minimax_h3_r2v)Each workflow includes a preview, Comfy Cloud / Download cards, input material cards, and prompting tips verified against the node schemas in
comfy_api_nodes/nodes_minimax.py(duration 5-15s, 2K resolution, reference tag conventions, image/video constraints).