HeyGen reference to video: 9 images, 3 videos, 3 audio

HeyGen's September 2026 changelog says reference_to_video on heygen-video-1 needs at least one image or video and accepts up to nine images, three videos and three audio recordings, twelve references in total. On Sume the same shape exists: up to 9 reference images, 3 reference videos and 3 audio files, with audio allowed only alongside an image or video.

HeyGen facts are from its changelog; Sume facts are from Video 1.0 and Video generation, both read 2026-09-30.

What are the reference limits on each side?

The per-field limits line up. The Sume docs do not state a combined cap of twelve, so check the fields rather than assuming the total.

Does every Sume video model honor audio references?

No. In the model catalog, a model accepts a reference type only if supported_input_references lists it. The docs say audio and video references are honored by the Seedance 2.x models, Wan 3.0, MiniMax H3 and MiniMax H3 Max. Check the GET /v1/videos/models catalog output for the model you plan to call before sending audio.

How do I send references to Sume?

On the Video 1.0 compatibility URL, pass the three URL arrays next to prompt. Use public HTTPS media URLs only. Video 1.0 is marked as retiring soon; for new work the docs point to POST /v1/videos with a catalog model id, where references go in input_references.

curl -X POST https://api.sume.com/v1/video-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A product turntable in the style of the references",
    "reference_image_urls": ["https://example.com/look-1.png"],
    "reference_video_urls": ["https://example.com/motion.mp4"],
    "reference_audio_urls": ["https://example.com/voice.mp3"]
  }'

What should I check before porting a HeyGen request?

Count each type separately, drop audio if you have no image or video, and confirm the target model lists the types you send. For a longer walkthrough of the Sume side, see reference-to-video with image, video and audio.