MOSS-Audio-8B-Thinking-ComfyUI

Audio understanding, transcription, captioning and question answering with text output.

This is a converted model package for Jack Spooner’s ComfyUI fork, derived from OpenMOSS-Team/MOSS-Audio-8B-Thinking. It is not a new fine-tune.

Required code

The integration modifies ComfyUI core; this is not a drop-in custom_nodes package.

• Model integration commit
• Implementation · Offline converter
• Development branch · Pinned compatible reference snapshot 9d9053c5e
• Installation, node usage and acceptance details

For an existing ComfyUI installation, use the linked core changes and their prerequisites against a compatible base. The integration commit alone is not asserted to be an independently cherry-pickable patch. The pinned reference snapshot includes the required shared loaders, attention and memory changes. Follow the fork’s installation README and dependencies when reproducing that snapshot; update the application only while its queue is idle. Upstream ComfyUI is Comfy-Org/ComfyUI; its README describes the underlying node engine and model management.

Files and installation

Place moss_audio_8b_thinking_bf16.safetensors in ComfyUI/models/text_encoders/. The tokenizer and required inference configuration are embedded; original HF shards are not required at inference.

File	Size
moss_audio_8b_thinking_bf16.safetensors	18.110 GB

Manifest/provenance files record the conversion and source revisions. SHA256SUMS and FILES.json describe this release. Filenames and model folders matter. No original source snapshot needs to be converted again.

Usage

Select moss_audio_8b_thinking_bf16.safetensors in Load CLIP, type moss_audio. Connect CLIP to Generate Text, and connect Load Audio to its audio input. Start with an explicit 512 or 2048 new-token budget. A caption prompt such as “Describe the foreground and background sounds in 3–5 sentences; state uncertainty” is more useful than a generic prompt. Select your own input audio in the supplied workflow.

Example graphs are in workflows.

Verification and limits

Full BF16 reference comparisons and RTX 5090 queue/cancellation/offload/reload tests are documented, including input coverage through two minutes. Instrument identification, exact timestamps and quiet-background detection can be wrong. Other accelerators and low-VRAM capacity are unverified. The 40,960-token context includes expanded audio and requested output.

These statements describe the linked retained tests, not a new cross-platform acceptance suite. File size is not a minimum RAM/VRAM requirement. No inference-time downloads or external inference worker are required by these native model paths.

Conversion and provenance

Streams 901 original BF16 tensors into native ComfyUI keys and embeds the original tokenizer as a byte tensor. No fine-tuning or new quantization is performed.

Source model revision: aacbb8036eb06dd968c105f91d2663af8fbee720. See PROVENANCE.json, the conversion manifest and NOTICE. No training data, private inputs or generated user recordings are included.

License and attribution

The source model is Apache-2.0. Retain LICENSE and NOTICE, and identify this repackaging as a modification.

See LICENSE, NOTICE, and the retained upstream model card/source notices. Model authors and publishers retain credit; this is a third-party ComfyUI repackaging and does not imply endorsement. The weights’ terms and the separately linked ComfyUI code license are distinct.

The example workflow files are copied/adapted from the linked ComfyUI fork; its GPL-3.0 license is retained in LICENSES/COMFYUI-GPL-3.0.txt. This does not relicense the model weights.